Every question about AI visibility. With the answer.

Generative engine optimization (GEO) is the work of getting a website to appear in the answers that AI systems write: as a source under the answer, as a name in the text, or both. The term comes from a research paper by researchers at Princeton University and IIT Delhi, posted online in November 2023 and published at the KDD conference in 2024. What appearing means exactly is not defined. AIGeschikt therefore counts two outcomes separately: mentioned, when your name appears as a whole word in the answer text, and cited, when a link in the source list points to your domain.

Also covered in What is GEO? · Mentioned or cited (in Dutch)

Generative engine optimization works for your website in two steps: an AI system with web search first has to pull a page from your site into its search results, and then decide to use that page as a source. In AIGeschikt's measurement from August 17 through September 16, 2026, a website appeared as a source in 52% of ChatGPT answers when one of its pages had been retrieved (81 of 157), and in 69% for Claude. If you are not retrieved, the work is about crawlers and findability. If you are retrieved but not cited, the work is about what is on your page.

Also covered in AI bot tracking · Diagnosis per buying question

Your website needs AI SEO when your buyers put their questions to AI assistants and you are missing from those answers while a competitor appears in them. You find out by asking ChatGPT, Perplexity and Google ten buying questions on three different days and noting for each answer whether your name comes up and whether your website is among the sources. AIGeschikt's rule of thumb: if you appear in at least half of the answers on every system, keeping an eye on it is probably enough. If you appear in fewer than one in five answers while a competitor appears often, you need AI SEO.

Also covered in Having your AI findability tested (in Dutch)

Answer Engine Optimization delivers five kinds of measurable outcomes: an impression in a Google AI feature, a citation as a source, a mention of your brand name, a referral that brings a visit, and a visit from an AI crawler. Each outcome has its own source. Impressions have appeared in the generative AI report in Search Console since August 31, 2026, and referrals in the AI assistant channel of Google Analytics 4. Citations and mentions in ChatGPT, Gemini, Perplexity and Claude you only see by asking questions yourself. The five measures cannot be added up into a single number.

Also covered in Measuring GEO marketing (in Dutch) · AI impressions in Search Console (in Dutch)

Answer Engine Optimization consists, in AIGeschikt's breakdown, of six activities: measuring, establishing why you are missing, creating, checking facts, publishing and measuring again. Measuring means putting fixed buying questions to AI assistants repeatedly; AIGeschikt measures 100 buying questions per business on ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, on several days a week and three times a month for Claude. The cause determines what gets made: a blog, a rewritten service page, structured data or an internal link. Every claim about the business is checked against recorded facts, publication follows only after the owner approves, and then the page goes into the measurement.

Also covered in How it works · Diagnosis per buying question

GEO, AEO and AI SEO are in practice names for the same work: being used by an AI system when it answers a question. Answer engine optimization is the older term and dates from the era of answer boxes and voice assistants. Generative engine optimization comes from a 2023 research paper on answers written by AI. AI SEO is the informal version. In the Netherlands, 'AI SEO' is searched 720 times a month and 'generative engine optimization' 590 times. The market has not yet agreed on the right name.

Also covered in What GEO actually means

An AI answer to a buying question uses a very different number of sources per system: in the buying questions of six Dutch businesses we measure, the week of October 5, 2026, an average of 2.7 for ChatGPT, 7.6 for Google AI Overviews and 4.5 for Perplexity. SISTRIX counted 11 sources per AI Overview internationally. Across all repetitions, one Dutch buying question on ChatGPT and Perplexity draws an average of 34.2 different sources, of which effectively 19.0 take part when weighted by use, in AIGeschikt's research into 214 questions in 30 markets. For comparison: the first page of Google shows ten results.

Also covered in Sources per assistant

A fixed position in AI answers does not exist: the list of sources and brands is put together anew for every question. In research by SparkToro and Gumshoe with 2,961 repeated queries, the chance that ChatGPT or Google's AI gave the same list of brands twice was smaller than 1 in 100. In AIGeschikt's own measurement from August 10 through September 14, 2026, a site that was cited three times in a week sat in the same place in the source list all three times in only 14% of cases. More stable is the share of answers you appear in, per AI system.

Also covered in AIGeschikt vs an SEO agency · Is GEO the same as SEO?

An AI answer changes a lot from week to week per question, while the total across many questions stays stable. At a Dutch security company that AIGeschikt measures, 44% of the places where it appeared as a source on one Monday or the next changed outcome (265 of 600 comparisons). The weekly total across 100 buying questions on four AI systems stayed between 28.7% and 30.9% over the same five weeks. In AIGeschikt's market research, the sources under a ChatGPT answer overlap week on week by 0.26. A change on one question therefore says little; a total that moves outside its margin does say something.

Also covered in Six stubborn assumptions · Measuring AI visibility

Generative engine optimization is largely the same discipline as SEO. Google writes that no extra requirements apply for AI Overviews and AI Mode: a page must be indexed and eligible to be shown with a text snippet. In the independent study C-SEO Bench (NeurIPS 2025), ranking a document higher in what the language model reads, which is what classic SEO does, improved its position on retail questions by 2.77 places, against 0.36 for the best rewriting technique. What does differ are the questions: which buying questions you are missing from, and which pages get retrieved in your place, you only see by asking those questions yourself.

Also covered in Different content for GEO? (in Dutch) · Is GEO the same as SEO?

GEO SEO looks beyond Google rankings at the sources AI assistants cite for your buying questions. In AIGeschikt's count of 320,681 source mentions for Dutch buying questions, from July 3 through September 7, 2026 on five AI systems, a business's own domain accounted for 2.2%. All mentions together came from 18,996 different domains, including direct competitors, comparison sites, organizations such as the Consumentenbond (the Dutch consumers' association) and the ANWB, and YouTube as the second most cited domain. In that count, being named on a comparison site or in an industry overview weighs as much as your own page.

Also covered in Who gets cited in the Netherlands

A high Google ranking does not automatically deliver GEO SEO. At a Dutch security company, the average Google position of 39 pages was almost unrelated to how often Google showed them in AI Overviews and AI Mode: the rank correlation was around zero. The best-ranking article, on average in the top five, was not cited as a source on any of the 100 measured buying questions in five weeks. The most frequently cited page, a comparison of the best options for home use, sits at position 15 on average. How often a page appeared in Google was clearly related to AI impressions.

Also covered in AIGeschikt vs an SEO agency · AI Overview sources outside the search results (in Dutch)

GEO SEO can affect your regular SEO results, because pages published for AI visibility sit on your own domain and are crawled by Google. Google writes that the same SEO fundamentals apply to AI Overviews and AI Mode as to regular search, so such a page usually helps in the regular search results too. That is not guaranteed: a new page on a topic you already have a page about can compete with it in Google. So have a new piece checked against your existing pages before it is published, for example by your SEO agency.

Also covered in AIGeschikt vs an SEO agency · Is GEO the same as SEO?

Existing SEO is only enough for AI SEO when AI search crawlers actually get into your site, your text is in the HTML without JavaScript, and there is a page that answers the buying question. Check access first: use curl to test which status code OAI-SearchBot, PerplexityBot and Claude-SearchBot receive, and look in your server log at what the real crawlers got. A firewall or CDN can refuse what robots.txt allows. In AIGeschikt's test on September 17, 2026 of 126 reachable, frequently cited .nl sites, robots.txt blocked those three crawlers from the homepage at not a single site, but the server refused a simulated AI crawler from OpenAI or Perplexity at 19 sites.

Also covered in AI bot tracking · When you need AI SEO (in Dutch)

Moving SEO budget to GEO is usually not necessary: leave the SEO budget in place and set a defined GEO portion next to it. Google writes that its AI features run on the same SEO foundation as regular search, and according to an Ahrefs study from February 2026 of 76,000 websites, Google sends 190 times more traffic to websites than ChatGPT. What SEO does not show is which buying questions AI assistants answer without you. Start with a free measurement: the AI report in Search Console, and ten buying questions on three days in ChatGPT, Perplexity and Google. Let the size of the GEO portion depend on the gap that measurement reveals.

Also covered in Splitting budget between GEO and SEO (in Dutch) · AIGeschikt vs an SEO agency

Split your marketing budget by leaving the SEO budget in place and setting a defined GEO portion next to it, with a measurement deciding whether it grows. A defensible start is roughly a fifth on top of your SEO spend. The agency Brainlabs suggests about 20% on top of SEO spend, or 7% to 10% of the digital marketing budget; Similarweb suggests about 25% of the search budget. None of those models has been studied for what it delivers. So fix a set list of buying questions at the start and assess the GEO portion per AI system after three months of measuring, not after one month.

Also covered in Moving budget from SEO to GEO

The sources of AI answers rank at the top of Google only in part. Ahrefs found in March 2026, across 4 million citations, that 37.1% of the pages cited in Google AI Overviews were in the top 10 and 36.7% outside the top 100, counting only the organic links. For AI assistants outside Google the overlap is smaller: in an Ahrefs study from August 2025 of fifteen thousand long-tail queries, 12% of the cited URLs were in Google's top 10, 8% for the in-text citations of ChatGPT and over a quarter for Perplexity.

Also covered in No top 10 needed for AI Overviews (in Dutch) · AI Overview sources outside the search results (in Dutch)

AI assistants send only a fraction of the visits that search engines send. According to Statcounter, in the Netherlands in August 2026, 94.1% of referral traffic to websites came from search engines, 5.72% from social media and 0.18% from AI chatbots. Conductor found, across 1,215 domains of its own enterprise clients, that 1.08% of all website traffic came from AI referrals. Those figures only count people who click through: a buyer who reads three company names in ChatGPT and then goes to one of those sites directly or googles it counts as direct traffic or as search traffic.

Also covered in Outcome measures for AI visibility (in Dutch) · Measuring GEO marketing (in Dutch)

Visitors from AI assistants do not demonstrably convert better than visitors from Google; the evidence is contradictory. Ahrefs measured on its own site that visitors from AI search created an account 23 times more often, and itself doubts whether that holds up. The largest published measurement, by Kaiser and Schulze in Marketing Science across 973 online stores from August 2024 to July 2025, found that organic search traffic converted 13% better than traffic from ChatGPT. So compare your own figures: Google Analytics 4 has had a separate channel for traffic from AI assistants since May 2026.

Also covered in Measuring GEO marketing (in Dutch)

A business gets into a ChatGPT recommendation in two ways: ChatGPT retrieves the business's own page, or it retrieves someone else's page that mentions the business. For the first route, your page must be findable through a sitemap or an internal link, your server must let OAI-SearchBot and ChatGPT-User in (a 200 in your server log, not a 403 or 429; GPTBot collects training material and does not count here), and the text must be in the HTML without JavaScript. For the second route, find the comparison sites, industry overviews and 'best X' lists that currently appear in the answers to your buying questions, and make sure you are named on them.

Also covered in What you can do about it in practice · The crawler check with curl (in Dutch)

The content formats with the best measured results for ChatGPT are 'best X' lists, pricing and product pages, and long pages with a lot of content per subheading. Across 750 questions and 26,283 source URLs, Ahrefs found that 'best X' lists were the largest page type at 43.8%. DecaGEO saw pricing pages that appeared in ChatGPT's search results get cited in 34.5% of cases, blogs in 6.1%. SE Ranking counted 5.1 citations for pages over 2,900 words against 3.2 for pages under 800 words. AIGeschikt found no controlled measurement for FAQ pages or original research, and nearly all of the evidence comes from parties with a commercial interest.

Also covered in Who gets cited in the Netherlands · Two pages, half of the citations (in Dutch)

AI visibility differs between assistants because each system picks different sources and shows a different number of them. In the buying questions of six Dutch businesses we measure, the week of October 5, 2026, an AI Overview showed an average of 7.6 sources per answer and ChatGPT 2.7. Of the sites cited for a question, 83% appeared with only one of ChatGPT, Gemini, Perplexity and Google AI Overviews. So always read an AI visibility figure per system, never as a single average.

Also covered in Six persistent assumptions · No AI visibility as a single number (in Dutch)

The parties that help you get mentioned in ChatGPT are mainly the websites ChatGPT consults as sources, and those are far from only providers. AIGeschikt counted 4,604 source mentions across 1,399 ChatGPT answers to Dutch buying questions about security (August 17 through September 16, 2026): 41.7% came from providers, 22.5% from comparison and consumer sites, and 19.8% from government, police and industry associations. The Consumentenbond (the Dutch consumer association) appeared in 21.5% of the answers with sources, the business's own site in 1.5%. A mention in a product test, on a review platform or with a quality mark therefore counts, although such a mention does not guarantee a place in the answer.

Also covered in Which sources AI assistants use (in Dutch) · Getting into ChatGPT recommendations (in Dutch)

An AI assistant that uses your website as a source usually does not mention your business name. In AIGeschikt's count of 17,168 answers from five AI systems (August 10 through September 11, 2026), a business appeared as a source 2,173 times, and in 74% of those its name did not appear in the text. For Perplexity that share was 86%, for Google AI Overviews 75% and for Gemini 53%. Conversely, a business was named in 34 of the 603 mentions without its own site being among the sources. If you only search for your name in answers, you see roughly a quarter of the times your site shapes the answer.

Also covered in Outcome measures for AI visibility (in Dutch)

For Dutch buying questions, AI assistants mostly use ordinary business websites: they receive 76.6% of all source mentions in AIGeschikt's research on ChatGPT and Perplexity (August 17 – September 27, 2026; for Perplexity the retrieved search results count). Consumentenbond, Trustoo and Trustpilot together account for 10.8%, of which consumentenbond.nl alone takes 3.9%. Wikipedia and Reddit, often named in international advice, are marginal in the Netherlands at 0.15% and 0.86% of mentions. Per site, a comparison site does get almost eight times the average number of mentions.

Also covered in The public measurement register · Which sources AI assistants use (in Dutch)

The number of sources per AI answer differs sharply between systems. In the buying questions of six Dutch businesses we measure, the week of October 5, 2026, Google AI Overviews showed an average of 7.6 sources per answer, Perplexity 4.5, Claude 4.0, Gemini 3.4 and ChatGPT 2.7. The shorter the list, the fewer places there are to win per answer. A high share in AI Overviews is therefore less remarkable than it looks, and a low share in ChatGPT less dramatic. So always judge a share next to the number of sources the system shows.

ChatGPT is by far the most used AI assistant in the Netherlands. According to Statcounter, ChatGPT accounted for 81.7% of referrals from AI assistants to websites in August 2026, Perplexity for 3.2%. Newcom's AI Monitor counts 7.2 million Dutch AI users, of whom 2.5 million use it almost daily. So weight any measurement of your AI visibility toward what your customers use, and read the ChatGPT figure as the most important one.

Also covered in No AI visibility as a single number (in Dutch) · When you need AI SEO (in Dutch)

A VLOSE (Very Large Online Search Engine) is a search engine with an average of at least 45 million monthly active users in the EU, designated by the European Commission and therefore subject to the strictest rules of the Digital Services Act. Since August 31, 2026, ChatGPT has been a VLOSE, alongside Google Search and Bing. Your own obligations as a business do not change: the rules apply to the search engine. The designation does mean that the EU treats ChatGPT as a search engine, with 159.1 million ChatGPT search users reported by OpenAI, between Google Search (364 million) and Bing (119 million).

ChatGPT's sources resemble Bing's search results more than Google's. In February 2025, in a sample of 100 searches, Seer Interactive found that 87% of the citations from ChatGPT's search feature matched Bing's top-20, against 56% for Google's search results. The matching Google pages also ranked low, at a median position of 17. A good Google ranking therefore does not automatically make you a source in ChatGPT. Registering with Bing Webmaster Tools and allowing OAI-SearchBot in robots.txt and in your firewall are separate steps for that reason.

Also covered in Checklist of GEO improvements (in Dutch) · What a VLOSE is (in Dutch)

AI Overviews are the summaries Google generates with AI and places at the top of the search results, with links to the websites the information comes from. In the Dutch interface the block is called 'AI-overzicht', and since May 20, 2025 AI Overviews have been available in Dutch. They do not appear for every search. The links in the block are a place where your website can be shown, above the regular results. AI Mode is something else: a separate search mode in which you hold a conversation with the search engine, whose Dutch rollout began on October 8, 2025.

Also covered in GEO, AI SEO or AEO · Measuring AI visibility

Google shows an AI Overview only when its systems determine that it adds something to the regular search results, and Google does not disclose which searches those are. In September 2025, across 146 million desktop search result pages, Ahrefs measured an AI Overview on 20.5%: on 9.5% of one-word searches, on 46.4% of searches of seven words or more, on 21.4% of informational and on 4.3% of commercial searches. In AIGeschikt's measurement of 548 fully written buying questions, an AI Overview appeared in 90% of the measurements. For your own search terms, you establish it by checking them on at least three different days.

Also covered in AI Overviews: what are they? (in Dutch)

Your website can appear in AI Overviews as soon as the page is indexed and eligible to be shown in Google with a text snippet; Google names no additional technical requirements. A place in the top 10 is not a condition. Ahrefs looked at 4 million addresses cited in AI Overviews from 863,000 search result pages: 37.1% ranked in the organic top 10, 26.2% at positions 11 to 100 and 36.7% outside the top 100. So check three things: Googlebot may fetch the page, the URL Inspection tool in Search Console shows it as indexed, and there is no nosnippet or max-snippet:0 on it.

Also covered in Is GEO the same as SEO? · AI Overview sources outside the search results (in Dutch)

You do not need to rebuild your content for AI Overviews: Google names no additional requirements and no special optimization. A good position in the regular results is not the only way in, though. On September 16, 2026, AIGeschikt compared 3,107 sources from 385 AI Overviews on mostly Dutch-language buying questions with the organic results on the same search page: 45% were not among them, and 41% came from a site that did not appear organically on that page at all. So besides your main pages, also answer the questions around them, such as what something costs, how long it takes and what the difference is between two options.

Also covered in Different content for GEO? (in Dutch) · No top 10 needed for AI Overviews (in Dutch)

Fewer clicks since AI Overviews are not necessarily caused by AI Overviews: first rule out measurement errors, seasonality, Google updates and lost rankings. When you compare with a period before April 28, 2026, compare only your clicks, because Search Console did not record impressions correctly from May 13, 2025 to April 27, 2026. Then put searches with and without an AI Overview, at roughly the same position, side by side: if the CTR drops only for the first group, AI Overviews are costing you clicks. Pew Research found among 900 American adults that 8% clicked a search result when an AI summary was shown, against 15% without one.

Also covered in AI Overviews: what are they? (in Dutch) · AI impressions in Search Console (in Dutch)

No single source shows clicks from AI Overviews separately; you explain a drop in clicks by combining four sources. Search Console counts clicks from AI Overviews and AI Mode under the Web search type, and since August 31, 2026 the report for generative AI shows only impressions per page, country, device and date. Since May 2026, Google Analytics 4 has an AI assistant channel ('AI-assistent' in the Dutch interface) covering ChatGPT, Gemini and Copilot among others, but it puts visits from AI Overviews under Organic Search. Bing Webmaster Tools counts citations in Copilot, and your server log shows which AI crawlers fetched your pages.

Also covered in AI impressions in Search Console (in Dutch) · Outcome measures for AI visibility (in Dutch)

Since August 31, 2026, you can exclude your website from AI Overviews with the Search generative AI setting under Settings in Search Console (called 'Generatieve AI in Zoeken' in the Dutch interface). That setting removes your site from AI Overviews, AI Mode and the AI features in Discover; according to Google, most content is excluded 1 to 2 days after the change, after which you no longer get traffic or impressions from those features. Google does not use the choice as a signal for the regular results. The nosnippet rule also works, but it removes your regular text snippet as well. The often-mentioned Google-Extended does not work here; that token governs training and Gemini.

Also covered in When Google shows AI Overviews (in Dutch) · Outcome measures for AI visibility (in Dutch)

A place in AI Overviews does not automatically give you a place in AI Mode: Google writes that the two can use different models and techniques, so the answers and the links shown also differ. AI Overviews appear among the regular search results as a short summary; AI Mode is a separate search mode in which you ask follow-up questions, whose Dutch rollout began on October 8, 2025. Search Console counts both together, including in the separate report for generative AI, so there you cannot see in which of the two your page was shown.

Also covered in AI impressions in Search Console (in Dutch)

An AI Overview is far from the same every time, even for the same search. Ahrefs tracked over 43,000 keywords and found that an AI Overview changed between two consecutive observations in 70% of cases, with an average overlap of 54.5% in the links. In AIGeschikt's measurement of 348 buying questions measured at least four times, Google always showed an AI Overview for 82%, never for 6% and sometimes but not always for 12%. Time of day, location, language, device and earlier search activity also play a role, so a single check says little.

Also covered in How much an AI answer changes week to week (in Dutch) · Lead times in GEO (in Dutch)

Pages cited as a source in an AI Overview get more clicks than pages that are not, but fewer than for searches without an AI Overview. Seer Interactive measured that cited pages got 120% more organic clicks per impression than uncited pages, although that number of clicks stayed 38% below the level of searches without an AI Overview. For informational searches in the third quarter of 2025, Seer saw a click-through rate of 0.70% for cited brands, against 0.52% for uncited ones. Seer itself writes that these figures do not prove the citation produces the clicks: strong brands may simply get cited more often.

Also covered in Fewer clicks since AI Overviews (in Dutch)

For GEO SEO you usually do not need different content than for regular SEO. Google states in its documentation that AI Overviews and AI Mode have no additional requirements and need no special optimization. In C-SEO Bench, an independent replication from 2025, classic SEO on shopping queries was 7.7 times more effective than the best-performing rewriting technique. What makes the difference is mainly the question your page answers and whether the page can be fetched at all.

Also covered in Is GEO the same as SEO?

Content that answers a buying decision worked best for GEO SEO in AIGeschikt's measurement. At one Dutch security company, AI systems cited 9 of the 21 blog articles as a source within five weeks, and most of those answer a buyer's choice, such as the best systems for the home (521 citations) or with or without a subscription (80). Ten general blogs about burglary prevention earned zero citations together on the same 100 buying questions. A page gets cited on the questions it answers; the topic alone is not enough.

Also covered in What you actually do about it · Content for ChatGPT (in Dutch)

You improve your AI SEO first with three conditions whose effect has been demonstrated: let the AI crawlers in, put your text in the HTML instead of loading it with JavaScript, and make sure your pages are indexed. After that, write pages for the buying questions you do not have a page for yet. Schema markup is not a good starting point: Ahrefs tracked 1,885 pages that received JSON-LD against 4,000 control pages and found no clear gain on any AI system.

Also covered in What you actually do about it · GEO improvement checklist (in Dutch)

You do not have to rewrite all your existing copy for GEO optimization; rewrite only the pages that miss a buying question. At three companies AIGeschikt measures, 806 of the 1,907 AI answers that cited their site (42%) included a page outside the blog, such as the homepage or a service page; per company that share ranged from 19% to 60%. A page that is already being cited is best left alone without a baseline measurement: you do not know which part of the text earns the citations.

Also covered in Automatic execution · Two pages, half the citations (in Dutch)

GEO optimization changes four things on a website: who may fetch the pages, which pages exist, what is on those pages, and how they are marked up, for example with schema.org or an llms.txt. Only for access do the platforms themselves document the effect: OpenAI writes that sites that block OAI-SearchBot do not appear in ChatGPT's search answers. In controlled studies, text changes made for AI helped less often than they hurt: in C-SEO Bench, adding statistics lowered a document's position in 19 of the 24 test setups.

Also covered in GEO improvement checklist (in Dutch) · Automatic execution

An llms.txt has no demonstrated effect on a website's AI visibility. SE Ranking studied about 300,000 domains and found no relationship between having an llms.txt and the chance of being cited by an AI system. Ahrefs analyzed the server logs of 137,210 domains: among the sites with an llms.txt, 97% of the files received not a single request in May 2026. No negative effect has been measured either, so an existing file can stay; just do not pay extra for one.

Also covered in What is GEO? · llms.txt: specification and use (in Dutch)

An llms.txt file gives an AI system, at /llms.txt, a short description of your site and a list of links to your most important pages, in markdown. The file is a proposal by Jeremy Howard of Answer.AI from September 2024, in version 2 since August 2026, and not a standard. An llms.txt does not control access: you allow or block crawlers in robots.txt or your firewall. In AIGeschikt's server log, recognized crawlers did not fetch /llms.txt a single time from September 3 through 17, 2026, and fetched /robots.txt 198 times.

Also covered in llms.txt: what it does (in Dutch)

Schema markup earns no extra AI citations according to the largest measurement with a control group. Ahrefs tracked 1,885 pages that received JSON-LD between August 2025 and March 2026, against 4,000 comparable pages without it. In Google AI Mode (+2.4%) and ChatGPT (+2.2%) the difference could not be distinguished from zero; in Google AI Overviews citations fell by 4.6%, small but statistically significant. These were pages that were already cited often. Existing schema can stay: Google uses structured data for rich results.

Also covered in GEO improvement checklist (in Dutch)

You check whether AI crawlers can get into your website with one curl command per crawler, and then in your server log or CDN dashboard. A 200 means the request succeeded; a 403 means the server refuses access, the code a firewall or security plugin returns most often. A 403 on an imitated crawler is not proof yet: in AIGeschikt's test of September 17, 2026, all 5 .nl sites that refused its imitated PerplexityBot still appeared as a source in Perplexity answers that month. Google Analytics does not show AI crawlers: Analytics runs on JavaScript, and the crawlers of OpenAI, Anthropic and Perplexity do not execute it.

Also covered in AI bot tracking

AI crawlers that search or fetch a page on behalf of a user should be allowed if you want to appear in AI answers: OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User and Googlebot. OpenAI writes that sites that block OAI-SearchBot do not appear in ChatGPT's search answers. GPTBot and ClaudeBot collect training material; according to the providers, blocking those two does not remove you from the search answers. Blocking Google-Extended, however, can keep you out of Gemini's answers.

Also covered in AI bot tracking · Outcome measures for AI visibility (in Dutch)

The crawlers of OpenAI, Anthropic and Perplexity do not read text that only appears on the page through JavaScript. Vercel and MERJ established from server logs at the end of 2024 that they do not execute JavaScript, across 569 million GPTBot requests and 370 million from Claude in one month. Google's crawler does execute JavaScript, so the problem weighs more heavily for ChatGPT, Claude and Perplexity than for AI Overviews. Search your page's source code for a sentence that is visible on screen: if you cannot find it, JavaScript is placing your text.

Also covered in When you need AI SEO (in Dutch)

A high GEO audit score says little about your AI visibility: the score shows whether you are below a technical threshold, not whether you get mentioned. Search Engine Land found, across 107,352 pages in AI Overviews and AI Mode, a correlation of −0.12 to −0.18 between load time and visibility, a weak relationship. Yellify found no relationship among 80 Dutch brands between its own GEO score and how often a brand was mentioned. Below the threshold, technology does count: a blocked crawler makes you invisible to that AI system.

Also covered in GEO improvement checklist (in Dutch) · What a GEO audit contains (in Dutch)

AI citations are not spread evenly across a site's pages: at one Dutch company AIGeschikt measured for five weeks, 888 of the 1,804 citations (49%) went to two of the 60 pages in the sitemap. Those two were a blog comparing the best systems for the home and the service page of the main product. Not a single citation went to 39 pages, including the frequently asked questions page. A skewed distribution becomes a risk if those few pages disappear or go out of date.

Also covered in Rewrite existing copy? (in Dutch)

You measure your own AI SEO by repeatedly asking ChatGPT, Perplexity and Google a fixed list of buying questions and counting, per AI system, in what share of the answers your website appears as a source. Look at that total rather than at individual questions: in AIGeschikt's measurement at one company, the weekly total stayed between 28.7% and 30.9% for five weeks, while 44% of the cited spots changed from Monday to Monday. For a first impression, take twenty buying questions, ask them on three different days, and note for each answer whether your site is among the sources.

Also covered in Measuring AI visibility · Getting your AI findability tested (in Dutch)

You measure AI findability on customer questions as the share of answers in which you appear, per AI system, across a fixed set of customer questions that is asked several times. An average position in AI answers is not a usable measure, because the list is put together anew for every question: in AIGeschikt's measurement, a site that was cited three times in one week sat in the same spot in the source list all three times in only 14% of cases. For every figure, ask how many answers it covers, whether the answers you are missing from are counted, and what margin comes with it.

Also covered in The measurement protocol · Measuring AI visibility

You can test your AI findability in six places: yourself with your own buying questions, for free in Google Search Console, Bing Webmaster Tools and Google Analytics 4, with a free online checker, with measurement software from €25 per month excluding VAT, with a one-off scan, or continuously as part of a service. The difference lies mainly in what gets measured: how many questions, on which AI systems, how often repeated, and whether the figure comes with a margin. None of the free reports shows whether you appear in ChatGPT on your buying questions. Prices read on September 15, 2026.

Also covered in Measuring AI visibility · Measuring GEO marketing (in Dutch)

You measure the effect of GEO marketing with a baseline and a control group: a fixed set of buying questions, each asked several times a week, counted separately per AI system, and split in advance into questions you work on and questions you deliberately leave alone. If only the first group rises, the effect is probably your work; a before-and-after comparison cannot tell a change in an AI system apart from your own work. Count on weeks: in AIGeschikt's measurement, OAI-SearchBot fetched 7 of 27 new blogs, with a median of 104 hours until the first visit, and a fetch is not yet a citation.

Also covered in What you actually do about it · Lead times in GEO (in Dutch)

An AI visibility figure needs a margin of error because an AI answer to the same question comes out differently every time, so a percentage based on few answers mainly shows that fluctuation. At a measured share of 30%, the true value lies with 95% confidence between 5% and 77% at 3 measurements, between 22% and 40% at 100 measurements, and between 27% and 33% at 1,200 measurements. If the margins of two measurements overlap widely, nothing has changed. In AIGeschikt's survey of September 7 and 8, 2026, one of eleven international providers published a margin of error, and none of five Dutch providers did.

Also covered in The measurement protocol · How much an AI answer changes per week (in Dutch)

Measuring the same questions every week is necessary because only then can you see whether your visibility changes: if you change the question list between two measurements, you are measuring the question list and not your visibility. Repetition is also necessary because an AI answer differs each time. In research by SparkToro and Gumshoe, with 2,961 repeated queries, the chance that ChatGPT or Google gave the same list of brands twice was smaller than 1 in 100. A rise only counts when the total across all your questions, per AI system, falls outside its margin.

Also covered in What you actually do about it · The measurement protocol

You see which AI crawlers visit your website in your server log or your CDN's logs, from the user agent of each request, such as GPTBot, OAI-SearchBot, PerplexityBot and ClaudeBot. Each crawler has its own purpose: at OpenAI, GPTBot fetches content that may be used for training, OAI-SearchBot serves ChatGPT's search features, and ChatGPT-User visits a page at a user's request. A crawler visit only says that a page was fetched, not that the page was cited.

Also covered in Outcome measures for AI visibility (in Dutch) · Fetched and cited (in Dutch)

Visits from ChatGPT show up in Google Analytics 4 in the AI Assistant channel (called 'AI-assistent' in the Dutch interface), which Google added in May 2026 for traffic from sources such as ChatGPT, Gemini, Deepseek, Copilot and Grok. ChatGPT also adds utm_source=chatgpt.com to its referral links itself. Two groups do not show up there: visits from Google's own AI Overviews and AI Mode, which fall under Organic Search, and people who read your name in an answer and look you up later, because they arrive as direct or organic traffic.

Also covered in Outcome measures for AI visibility (in Dutch) · When you need AI SEO (in Dutch)

For AI visibility you can measure five outcomes: a mention of your name in the answer text, a citation of your page in the source list, an impression in AI Overviews or AI Mode, a referral that brings a visit, and a visit from an AI crawler to your server. Each outcome has its own source: Search Console counts impressions, Bing Webmaster Tools counts citations in Copilot, Google Analytics 4 counts referrals and your server log counts crawler visits. Mentions and citations in ChatGPT, Gemini, Perplexity and Claude can only be established by asking the questions yourself. The five outcomes cannot be added up into one number.

Also covered in Measuring GEO marketing (in Dutch) · Getting your AI findability tested (in Dutch)

Which GEO tool fits best depends on four things: how many buying questions you want to track, on how many AI systems, whether you only want to measure or also want the work done, and your budget. If fewer than five buying questions really matter to you, you can check them yourself faster than you can sign up for a subscription; between twenty-five and a hundred questions, an entry-level subscription starts to make sense. Ask every provider how often it asks the same question: according to Schulte and colleagues (April 2026), the standard error of a per-brand figure only drops below 0.10 at seven repetitions per question per day. A measurement tool does not change anything on your site by itself.

Also covered in All comparisons · Choosing a provider for AEO (in Dutch)

You do not need a tool to choose between GEO and SEO, because that choice largely does not exist: according to Google, a page only has to be indexed and eligible for a snippet to appear as a source in AI Overviews or AI Mode, with no additional requirements. In C-SEO Bench (2025, over 1,900 questions), placing a document higher in the context, which is what classic SEO does, was 7.7 times more effective on shopping questions than the best-performing rewriting technique. What a measurement does add is showing for which buying questions you are missing from AI answers and which page appears there in your place.

Also covered in Is GEO the same as SEO? · Different content for GEO? (in Dutch)

A tool that combines GEO optimization with content changes can be recognized by five functions next to the measurement: content generation, publishing through a connection to your CMS, an approval step before anything goes live, editing existing pages, and checking figures against a source. Measurement, content production, crawler logging and competitor comparison are four different functions that are often sold under one name. So check for each function whether it is really there, for example by a number of articles per month in the offer, and ask whether the text is actually placed on your site or delivered as a document.

Also covered in Automatic execution · All comparisons

You do not need a tool to assess whether llms.txt is necessary for AI visibility: Google states in its AI optimization guide (updated July 10, 2026) that Google Search does not use such files. If you still want to check whether the file is set up correctly, Chrome Lighthouse does that for free and marks a missing file as Not Applicable. Semrush Site Audit reports a missing or badly formatted file as a Notice, the lowest level, and the free validator 37audits marks it as FAIL. Ahrefs found across 137,210 domains that 97% of llms.txt files were not fetched by anyone in May 2026.

Also covered in llms.txt: what it does (in Dutch) · llms.txt: specification and use (in Dutch)

You can analyze your visibility in AI Overviews for free in Google Search Console, which since August 31, 2026 has a separate generative AI report for every site, covering AI Overviews and AI Mode. The report shows impressions per page, country, device and date, but no clicks, no search queries and no position. If you want to know for which question you appear in an AI Overview and who appears next to you, you need a tool that runs searches itself and retrieves the AI Overviews. The two measurements complement each other.

Also covered in Measuring AI visibility · AI Overviews: what are they? (in Dutch)

Monitoring of AI mentions is offered by measurement software and by services that measure, and the choice mainly depends on whether a provider counts mentions and citations separately. The two rarely coincide: in AIGeschikt's count of 17,168 answers from five AI systems (August 10 through September 11, 2026), a company was not named in 74% of the 2,173 times its website was listed as a source. A tool that only searches for your name therefore sees roughly a quarter of the times your website serves as a source. Also ask whether the provider reports per AI system and publishes a margin of error.

Also covered in Measuring AI visibility · The measurement protocol

A GEO tool can point out improvements, but only three issues are known to make your page invisible when they fail: an AI crawler that cannot get in, text that only appears on the page after JavaScript runs, and a page that is not indexed in Google. For other issues that audits often report, no effect has been measured: adding structured data produced no gain in an Ahrefs measurement of 1,885 pages that were already cited often, and Yellify found no link between a GEO score and mentions across 80 Dutch brands. A tool that only crawls your site cannot see for which buying questions you are missing a page.

Also covered in What a GEO audit contains (in Dutch) · An audit score is a threshold (in Dutch)

Almost any measurement tool can measure your brand's AI visibility, but choose one that shows the result per AI system and do not base decisions on a single combined score. ChatGPT, Perplexity and Google AI Overviews use such different sources that an average mostly hides where you are missing: in AIGeschikt's measurement of 393 buying questions on which all three gave sources (August 17 to September 16, 2026), 89.3% of the combinations of question and cited address appeared in only one system. Also check whether every percentage states how many answers it is based on.

Also covered in Six persistent assumptions · Five AI systems differ widely (in Dutch)

Four tools measure your visibility in AI answers alongside your Google rankings: Semrush One, Ahrefs, SE Ranking and Nightwatch. On September 17, 2026 the entry plans cost €190 per month (Semrush One Starter, 50 prompts), €119 (Ahrefs Lite, 5 prompts), $129 (SE Ranking Core, 100 prompts) and €79 with annual billing (Nightwatch Starter, 50 prompts). None of the four says how often each prompt is asked per run or how certain the result is. Of the fourteen tools AIGeschikt reviewed, only Evertune publishes a margin of error, and Evertune does not track Google rankings.

Also covered in Choosing a provider for AEO (in Dutch) · The margin of error on every percentage (in Dutch)

To measure AI findability per customer question, you use a measurement tool, an SEO suite with AI tracking, or a service that repeatedly asks AI systems a fixed list of customer questions and records per question whether you are named or cited. Dutch measurement tools start at €25 to €49 per month excluding VAT, international ones at $29 per month (Otterly.AI Lite, 15 prompts). Google Search Console and Bing Webmaster Tools are free, but they count per page and not per question. Choose on the number of questions, how often each question is asked and whether you see the result per AI system. Prices read on September 17, 2026.

Also covered in Fourteen tools side by side (in Dutch) · Measuring AI visibility

Promptwatch costs $95, $245 or $579 per month, in euros €85, €205 or €499, checked on September 25, 2026 on promptwatch.com. Essential gives 1 seat and 50 prompts, Professional 2 seats and 150 prompts, and Business 5 seats and 350 prompts; every plan measures four AI models daily at the same time, out of eleven that Promptwatch supports. Professional also includes 1,500 agent credits, a knowledge base of 100 items that you fill yourself, and per month 5 AEO Articles and 5 content optimizations as drafts. The WordPress integration is listed as ‘coming soon’ on all three plans.

Also covered in All comparisons

Profound no longer lists a monthly price for brands on its pricing page, checked on September 25, 2026: there is a free seven-day trial, and after that only Enterprise, priced on request. In the trial, 50 prompts run daily on ChatGPT, Gemini and Google AI Overviews, in one language and one region, without history or export. Enterprise measures up to nine answer engines. Agencies buy a workspace per client for $399 per month. On September 7, 2026 the pricing page still listed Starter for brands at $99 and Growth at $399 per month, billed annually.

Also covered in All comparisons

Peec AI costs €70 to €425 per month, depending on plan and billing term, checked on October 7, 2026: Starter €70 with annual billing or €85 monthly, Pro €180 or €205, Advanced €360 or €425, and Enterprise custom. The API is only included in Enterprise; on September 17, 2026 Starter measured 50 prompts on three AI models of your choice. Users and languages are unlimited on every plan. A CMS integration is not in the feature list; WordPress appears only as a source for crawler logs.

Also covered in All comparisons · Fourteen tools side by side (in Dutch)

Scrunch costs $250 per month for the Core plan; Enterprise is custom-priced, checked on September 25, 2026. Core measures 125 prompts on ChatGPT, Perplexity, Google AI Overviews and Copilot, for one country and one language, with five licenses and five site audits per month. In terms of execution, Core includes one page optimization per month. According to the documentation AIGeschikt read on September 17, 2026, prompts are measured daily for the first 14 days and every 72 hours after that, and Scrunch shows how many answers a percentage is based on, without a margin of error.

Also covered in All comparisons · Fourteen tools side by side (in Dutch)

Promptwatch, Profound, Peec AI and Scrunch can all four measure Dutch prompts, but with different limits, checked on September 25, 2026. Promptwatch supports Dutch as a language and Peec AI has unlimited languages on every plan. Scrunch Core measures in one language, so if you choose Dutch, that is the only one. Profound measures one language and one region in the free trial; custom languages and regions are only available on Enterprise, priced on request. The interface and support of Profound, Peec AI and Scrunch are in English. Which sources carry Dutch AI answers is separately measured work.

Also covered in Who gets cited in the Netherlands · AIGeschikt vs Profound

GEO SEO in the Netherlands costs from €49 per month for measurement software only to €1,195 per month plus €2,495 in setup fees for full service, excluding VAT, as read on the pricing pages of six providers on September 8 and 9, 2026. A broader count of 14 providers on September 17, 2026 ranged from €25 to over €2,000 per month. That gap between the low and the high end is mainly a difference in product: measurement software changes nothing on your site, a content agency writes articles, and a managed service measures and executes. With measurement software, count your own hours too: industry estimates come to ten to twenty hours per month.

Also covered in What an AEO engagement costs (in Dutch) · AIGeschikt vs doing it yourself

Having AI SEO done in the Netherlands costs a median of about €500 per month at entry level; having it measured only costs a median of about €45 per month. Those medians come from AIGeschikt's count of the entry prices of 14 Dutch providers, read on September 17, 2026. Among the services that do the work, the entry price ranged from €95 to €695 per month, and the highest published monthly price was €2,295. The price gap between measuring and executing lies in the work: a measurement tool changes nothing on your site.

Also covered in What an AEO engagement costs (in Dutch) · Who carries out GEO optimization (in Dutch)

An Answer Engine Optimization engagement in the Netherlands costs from €25 to over €2,000 per month, depending on whether you only have it measured or have the work done. On top of that there may be a scan (free up to €295), a defined project (up to €1,500) or setup fees. So calculate the first year: with a service that does the work at around the median of about €500 per month, that comes to around €6,000, and with OpsCode for a service provider (€1,195 per month plus €2,495 setup) it is €16,835. Prices read on September 17, 2026.

Also covered in Buying AEO as a standalone service (in Dutch) · Who carries out GEO optimization (in Dutch)

Having your AI visibility improved in the Netherlands costs from about €25 per month for a measurement tool to over €2,000 per month for an agency, excluding VAT, as read on September 17, 2026. With a measurement tool you see where you stand and do the work yourself. Agencies that do the work charge more: NEXTRIQ charges €495 to €1,950 per month, OpsCode up to €2,295 plus €1,495 to €3,995 in setup fees. AIGeschikt measures, writes and publishes after your approval for €299 or €499 per month excluding VAT, with no setup fees.

Also covered in Pricing · What an AEO engagement costs (in Dutch)

A measurement of your AI findability costs from about €25 per month with standalone measurement software in the Netherlands; at AIGeschikt the measurement is included in every subscription, from €299 per month excluding VAT, and it cannot be bought separately. That measurement covers 1,200 measurements per week across 100 fixed buying questions, reported separately per AI system. For the percentage per buying question, the report states the margin within which the true value lies with 95% certainty. Internationally, Otterly.ai starts at $29 per month with monthly billing, for 15 questions (September 2026).

Also covered in Pricing · Measuring AI visibility

Answer Engine Optimization can be bought as a standalone service in the Netherlands, in four forms: measurement software, a one-off scan, a defined project or an ongoing service. Measurement software starts at €25 per month, a scan at Citeerbaar costs €295 and a Sprint of three to six weeks €1,495, excluding VAT (as of September 11, 2026). The form determines who does the work: with software and a scan, you carry out everything yourself. Read the terms too: at three of the eight providers examined, the pricing page and the terms said different things.

Also covered in What an AEO engagement costs (in Dutch) · Who carries out GEO optimization (in Dutch)

Carrying out GEO optimization takes hours to weeks per step: measured crawl times range from hours to days, and a first citation in an AI answer follows after one day to several weeks. In AIGeschikt's server logs of 27 new pages (September 2026), Googlebot visited after a median of 26 hours, PerplexityBot after 29 hours and OAI-SearchBot after 104 hours. Providers usually promise 4 to 12 weeks for first citations, without backing that timeframe with a measurement.

Also covered in Measuring GEO marketing (in Dutch) · AI bot tracking

A GEO audit in the Netherlands costs from free to €295 excluding VAT, as read on the pages of four providers on September 17, 2026. A free audit usually leads to a paid plan; a paid scan gives more explanation, but does not carry anything out either. The work after the audit costs extra: a defined project costs €695 to €1,500 excluding VAT. Ask in advance how often each question in the audit is asked, because a single measurement moment is only a rough indication.

Also covered in Who carries out GEO optimization (in Dutch) · What a GEO audit contains (in Dutch)

Most Dutch GEO providers charge no setup fees according to their own pages, but there are exceptions. OpsCode charges €1,495 to €3,995 in setup fees on top of the monthly amount, depending on the industry, and NEXTRIQ recommends starting with a one-off Fundament package of €1,500 excluding VAT. Also watch the contract term: OpsCode states six months fixed, and at Joogi an agreement runs for a year according to the terms, unless agreed otherwise. Prices and terms read on September 17, 2026.

Also covered in Pricing · Between the tools and the agencies (in Dutch)

A citation in AI answers disappears within a few weeks in half of the cases, according to published measurements. Scrunch, itself a vendor of measurement software, found across 3.5 million citation events a half-life of about 4.5 weeks: 3.4 weeks in ChatGPT and 5.8 weeks in Perplexity. In AIGeschikt's measurement at four companies, a site that was cited for a question in one week was cited again in the same AI system the following week in 78.3% of cases, and four weeks later in 67.9%.

Also covered in How much an AI answer changes per week (in Dutch) · Six persistent assumptions

You choose an Answer Engine Optimization provider on one question: how often do you measure the same question, and what is your margin of error? In AIGeschikt's inventory of September 7, 2026, one of eleven international providers published a margin of error; among five Dutch providers, on September 8 the count was zero of five. That question matters because AI answers differ each time: in a study by SparkToro and Gumshoe with 2,961 repeated queries, the chance of getting the same list of brands twice was smaller than 1 in 100 for ChatGPT and Google. Then ask whether the provider reports per AI system and works with a fixed question set and a baseline measurement.

Also covered in Choosing a GEO tool (in Dutch) · The margin of error behind every percentage (in Dutch)

You choose a suitable GEO agency on four points: how it measures, whether anything goes live without your approval, how long you are locked in, and who owns the content. Ask how many buying questions the agency measures, how often, on which AI systems and with what margin of error; on September 8, 2026, none of five Dutch providers examined published a margin of error. Get the term and notice period confirmed in writing, because the maximum notice period of one month that protects consumers does not apply to businesses. And under Article 2 of the Dutch Copyright Act (Auteurswet), copyright only transfers in writing, so ownership of the texts belongs in the agreement.

Also covered in Buying AEO as a standalone service (in Dutch) · Choosing a GEO specialist (in Dutch)

You choose the right GEO specialist with five questions: how do you measure, what changes on my site and who does it, does anything go live without my approval, how long am I locked in, and who owns the content. A good answer is concrete: a number of questions, a frequency and the AI systems by name, a number of pieces per month, and a contract term that matches on the pricing page and in the terms and conditions. In AIGeschikt's comparison of September 17, 2026, no provider examined answered all five questions on its own site. So have the answers written into the quote.

Also covered in Choosing a GEO agency (in Dutch) · Choosing a provider for AEO (in Dutch)

A GEO agency can take over five tasks: measuring whether you appear in AI answers, finding out why you are missing, writing or revising pages, publishing on your own site and then measuring again. Most Dutch offerings deliver one or two of those steps. Measurement tools, from €25 per month excluding VAT, only measure; content services write but do not always measure; a scan measures once and gives advice. Technical SEO, page speed, backlinks and local listings are usually not part of what a GEO service takes over. Prices read on September 17, 2026.

Also covered in How it works · The work is done for you (in Dutch)

Your current SEO partner can probably do AI SEO as well: Google writes that AI Overviews and AI Mode have no extra requirements on top of the regular SEO basics. Agencies do differ in how they show whether it works. Of eleven Dutch SEO agencies with a GEO or AI SEO service, on September 17, 2026 two described how they measure AI visibility, and none mentioned a margin of error. Ask your partner how many buying questions they measure, on which AI systems and how often, and whether they also check Bing, the AI crawlers and text that only appears after JavaScript runs.

Also covered in AIGeschikt vs an SEO agency · The crawler check with curl (in Dutch)

GEO optimization for websites is carried out in the Netherlands by six types of providers: specialized GEO agencies, SEO agencies that added GEO, content and automation services, measurement tools, freelancers and one-off scans. They deliver different things: agencies and freelancers do the work, content services write and publish, measurement tools and scans show where you stand. On September 17, 2026, published prices ranged from free to over €2,000 per month, some excluding VAT and some without any mention of VAT. Freelancers in the examples charged €12.50 to €80 per hour. First decide which part of the work you want to do yourself, and only then choose a provider.

Also covered in All comparisons · What an AEO project costs (in Dutch)

You can outsource generative engine optimization, but what you get depends on the form. A scan gives advice, a measurement tool gives insight, a content service delivers articles, and an agency or managed service does the work: measuring, determining what is missing, writing and measuring again. On September 17, 2026, Dutch prices ranged from a free baseline measurement or a €50 scan to monthly packages of over €2,000, excluding VAT. For each provider, check the measurement, the execution, the approval before publication and the contract, and calculate the cost of the first year instead of only the monthly fee.

Also covered in Buying AEO as a standalone service (in Dutch) · AIGeschikt vs doing it yourself

You can have a GEO audit done by a specialized GEO agency, a measurement tool or an SEO tool; on September 17, 2026, that cost from free to €295 excluding VAT in the Netherlands. Choose on what the audit delivers, not on the score: across 107,352 pages the correlation between load time and AI visibility was only −0.12 to −0.18. A useful audit checks per crawler whether OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot get in, whether your text is in the raw HTML, and on which buying questions you are not mentioned.

Also covered in Having your AI findability tested (in Dutch) · The crawler check with curl (in Dutch)

A GEO audit in practice contains up to twenty different points, from crawler access, text in the HTML and structured data to llms.txt, page speed and heading structure, plus a measurement of your mentions. In AIGeschikt's count of 16 audits and checklists, as of September 17, 2026, 2 of those 20 points have a demonstrated effect on AI visibility: AI crawler access and text in the raw HTML, both of them thresholds. Structured data, named by 11 of the 16 sources, has been tested without a positive effect: Ahrefs found no measurable difference in ChatGPT citations across 1,885 pages with new schema.

Also covered in Checklist of GEO improvements (in Dutch) · Schema does not earn AI citations (in Dutch)

After a GEO audit you usually get three things: a score or verdict per component, a list of improvements and a snapshot of how often you appear in AI answers. You do not get execution: an audit points things out but changes nothing on your site, and at Citeerbaar the implementation is a separate package. Read a total score with caution: Yellify found no link across 80 brands between its technical GEO score and how often a brand was mentioned. Start with the two thresholds that have a demonstrated effect: access for AI crawlers and text in the HTML.

Also covered in What a GEO audit contains (in Dutch) · An audit score is a threshold (in Dutch)

A measurement tool delivers the number; a managed GEO service delivers the number plus the intervention. With a measurement tool, the work that makes the difference stays with you: choosing topics, writing, rewriting and publishing. On September 17, 2026, Dutch measurement tools started at €25 to €49 per month excluding VAT; add the hours of whoever does the work, because those appear on no invoice. A measurement tool fits if you already have a marketing department or agency that handles the execution.

Also covered in Automatic execution · Choosing a GEO agency (in Dutch)

Handling AI visibility yourself is possible and costs no money, but it does cost time: according to agency rules of thumb, ten to twenty hours per month, plus forty to eighty hours of getting up to speed in the first months. For each piece you choose a topic, write, check the facts, publish and later check whether it had an effect; on top of that you ask the same questions again every week. Choosing is the hardest part: in AIGeschikt's research a Dutch buying question draws on average 34.2 different sources and a ChatGPT answer shows 3 to 4 of them, so without repeated measurement you cannot see which sources carry the answer.

Also covered in How an AI answer comes about · The pricing guide (in Dutch)

A GEO agency cannot guarantee a result in AI answers: those answers depend on factors outside any provider's control and differ each time you ask the same question. The research does not support such a promise either. The original GEO study (KDD 2024) found up to about 40% more visibility on its own benchmark, but C-SEO Bench (NeurIPS 2025) found that most text interventions largely had no effect. A promise of a fixed number of mentions within a fixed period therefore cannot be verified.

Also covered in Lead times in GEO (in Dutch) · Different content for GEO? (in Dutch)

AIGeschikt is a fully managed AI visibility service for small and medium-sized businesses in the Netherlands. AIGeschikt measures how AI assistants talk about your business, works out why you are not being mentioned, and carries out the improvements itself: articles, rewritten pages and schema.org, on your own site and only after you approve. The measurement runs across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. AIGeschikt is not a self-service dashboard you have to learn to operate, and it does not do technical SEO, link building or local search.

Also covered in The work is done for you (in Dutch) · AIGeschikt vs an SEO agency

AIGeschikt is for Dutch small business owners without their own marketing department who want AI assistants to name their business when someone asks a buying question, and for agencies that deliver this white-label to their own clients. The service fits when nobody in your business takes on the writing and publishing work on a regular basis. It does not fit if you want to pick the topics and write yourself: then you pay for work you end up redoing, and a measurement tool makes more sense. Agencies use the Partner package, with 3 clients included.

Also covered in For agencies · The work is done for you (in Dutch)

AIGeschikt costs €299 per month for Kern and €499 per month for Groei, excluding VAT. Kern delivers 3 finished pieces per month, Groei 5; both packages cover one domain. Agencies take Partner: €899 per month for 3 clients, plus €149 per additional client, also excluding VAT. Every package can be canceled monthly.

Also covered in All comparisons · Between the tools and the agencies (in Dutch)

AIGeschikt charges no setup fees: there is no onboarding fee and no extra work billed on top. You pay one fixed amount per month with everything included: €299 for Kern or €499 for Groei, excluding VAT. Some GEO agencies do charge a one-time startup fee, and that weighs heavily in the first year. So compare providers on the cost over twelve months, not just on the monthly price.

Also covered in What an AEO project costs (in Dutch)

AIGeschikt can be canceled monthly and has no fixed contract term. You can also switch in between: move from Kern to Groei or back at any time, and your measurement history is kept. Every week, the measurement uses the same 100 buying questions to show whether the work is paying off. A flat line is just as visible as a rising one, and AIGeschikt does not hold you to something that is not moving.

Also covered in Buying AEO as a standalone service (in Dutch)

Kern delivers 3 finished pieces per month, Groei 5. Groei also builds a new page in your own house style every quarter, runs a technical AI check of your site every quarter, delivers a quarterly report and gives priority in support, with an answer within one working day. A dedicated strategist, a monthly overview and a quarterly session belong to Enterprise. The measurement is the same in Kern and Groei: 100 buying questions and 1,200 measurements per week across 5 AI systems.

Also covered in The work is done for you (in Dutch)

With AIGeschikt, everything is included in the monthly price: measuring, diagnosis, writing, rewriting, schema and publishing. There are no hourly rates and no separate charges per article. On Kern you get 3 finished pieces per month, plus the page optimizations and schema the measurement calls for. That choice is deliberate: as soon as execution is billed per piece, the conversation turns to volume instead of results.

Also covered in For SMEs · The work is done for you (in Dutch)

At AIGeschikt, one piece is a finished intervention, not a text you still have to put online yourself. Each piece is chosen based on the weekly measurement, checked against your confirmed facts, given schema.org markup, published on your own site after you approve, and then tracked in the measurement. Kern delivers 3 per month, Groei 5. On top of that come the optimizations the measurement calls for, such as a rewritten page, extra schema or an internal link.

Also covered in The work is done for you (in Dutch) · Between the tools and the agencies (in Dutch)

AIGeschikt's first proposals are usually ready within a week. Visibility in AI answers then builds up over weeks: AI models do not pick up new content immediately, and the first meaningful shift usually shows only after a few measurement weeks. A single outlying measurement is noise, so drawing a conclusion after two weeks is drawing it too early. You follow the trend in your dashboard and in the weekly email.

Also covered in Pricing · Lead times in GEO (in Dutch)

AIGeschikt takes minutes of your time per week, not hours. On Monday you get one email with what is ready and why; you read it in two minutes and decide per piece whether it can go live. Now and then you also answer a question only you can answer, such as a price, a working method or a detail from your trade. Measuring, researching, writing and checking happen without you, and no technical knowledge is needed.

Also covered in How it works

AIGeschikt puts nothing live without your approval. Every proposal, whether it is an article, a page change or schema.org, first sits ready as a draft with the full text. Only after you click approve does AIGeschikt put it on your site. That approval is a hard rule, not a setting you have to switch on: there is no path by which content is published without your click. You approve with one click from your email.

Also covered in Automatic execution

A rejected AIGeschikt proposal does not go live and does not come back in that form. If you give a reason, AIGeschikt remembers it, so the next proposals match what you want more closely. If AIGeschikt gets something about your trade wrong, correct it as well: an article about your trade is only as good as the expertise in it. If you approved something earlier that you no longer like, AIGeschikt removes it again with one action.

Also covered in Automatic execution · For SMEs

AIGeschikt publishes on WordPress, Ghost, Shopify, Webflow, Wix, Squarespace and through custom APIs. The work always lands on your own site, first as a draft until you click approve. Connecting is fastest on WordPress, because there the connection is a plugin: through AIGeschikt's connector, drafts are placed directly in your own WordPress. Other platforms are connected through their own integration.

Also covered in For agencies · How it works

AIGeschikt also works for multilingual websites. If your site has Dutch, English and German pages, the same improvement lands on every language version, so you are equally strong in each language. The approval rule applies to the other language versions too: nothing goes live without your approval.

Everything AIGeschikt places on your site can be fully removed again with one action, at any time and at your request. So you are never locked in: what is live is there because you approved it, and what you no longer want, AIGeschikt takes down. That also applies if you cancel; AIGeschikt can be canceled monthly and has no fixed contract term.

Also covered in Pricing · Buying AEO as a standalone service (in Dutch)

AIGeschikt does not write generic AI text. Every piece is built on your own figures, quotes and trade vocabulary from onboarding, in your tone, and goes through a quality check that filters out generic AI language. The topic follows from the weekly measurement of 100 buying questions, not from gut feeling. For high-risk sectors, such as legal, medical and financial services, AIGeschikt adds the right disclaimers automatically.

AIGeschikt prevents made-up facts in two ways. It writes only with facts you have confirmed; those are recorded in your fact base, the Brain (called 'Brein' in the Dutch interface), and every claim is checked against it before anything goes live. A second AI model then strictly assesses the piece on factual accuracy and originality, among other things. If it does not meet the bar, AIGeschikt rewrites it or does not send it. You then read the full text before you approve.

Also covered in Pricing · The work is done for you (in Dutch)

AIGeschikt measures 5 AI systems: ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. For each system AIGeschikt uses the current model with live web search, so it measures what a customer would also see today. Google AI Overviews is included on purpose: it is not a chatbot but the AI answer at the top of the regular search results, and that is often where your customer lands first. Results are reported per system, not merged into a single number.

Also covered in Five AI systems differ widely (in Dutch) · No AI visibility as a single number (in Dutch)

AIGeschikt runs 1,200 measurements per week per domain. Those are not 1,200 different questions: your set holds 100 buying questions, and every question goes to ChatGPT, Gemini, Perplexity and Google AI Overviews on Monday, Wednesday and Friday, twelve times a week in total. Claude gets the same questions three times a month. Per month that comes to 5,000+ measurements. One measurement is a snapshot, because AI answers differ each time; only by repeating the same question do you get a percentage you can act on.

Also covered in The margin of error in every percentage (in Dutch)

AIGeschikt measures the same 100 buying questions every week because that is the only way to see a real trend. The question set is frozen: a list that keeps shifting produces a new baseline instead of a change. AIGeschikt measures on Monday, Wednesday and Friday, three states of the web rather than three snapshots of the same hour. The questions are asked the way your customer asks them, without your brand name in them; if the AI asks a follow-up question, AIGeschikt answers once, neutrally.

Also covered in How much an AI answer changes week to week (in Dutch) · The margin of error in every percentage (in Dutch)

AIGeschikt measures your competitors along with you. For each AI answer it records whether your business is mentioned and in which position among the providers (shown as a distribution, not as one average position), with what sentiment, and which competitors are mentioned instead of you or next to you. That way you see per buying question who gets the AI answers in your market. On the Groei package AIGeschikt steps in right away as soon as a competitor rises; on Kern that happens in the weekly round.

Also covered in Pricing

AIGeschikt's free AI scan asks ChatGPT and Google AI Overviews ten buying questions from your market, in Dutch or in English, whichever your customers use, each question three times: 60 answers in total. Within a few minutes you see for which questions your business is already mentioned, and who appears in the answers where you are missing. No credit card is needed and there is no obligation. If you then book a results call, AIGeschikt measures the same questions on Gemini, Perplexity and Claude before that call.

Also covered in Having your AI findability tested (in Dutch)

AIGeschikt's diagnosis finds the cause for each buying question where the weekly measurement shows you are missing. There are three possible causes: the answer is not on your site, there is no proof you can deliver it, or a competitor is stronger. AIGeschikt looks at your page as raw HTML, the way an AI crawler sees it. The cause determines the intervention: an article, a rewritten page or a technical change. AIGeschikt does not use a technical audit score as a predictor: whether a crawler can read your page is a threshold, not a lever.

AIGeschikt does not always write a blog post: the diagnosis decides the intervention. If information is missing, AIGeschikt writes an article. If the page already exists but does not convince, AIGeschikt rewrites it. Sometimes the intervention is smaller: extra schema.org on an existing page, or an internal link that makes a page reachable. Such an optimization does not count as one of the 3 pieces in Kern, but comes on top of them, at no extra cost.

Also covered in The work is done for you (in Dutch)

AIGeschikt shows per page which AI crawlers visit your site and when, such as GPTBot, ClaudeBot, PerplexityBot and ChatGPT-User. Each visit is sorted into one of three types: training, search or a live question from a user. The visits are read through AIGeschikt's WordPress connector, which recognizes and verifies crawlers by their signature, or through Cloudflare if your site sits behind it. AI bot tracking is included in all three packages.

Also covered in Outcome measures for AI visibility (in Dutch)

Agencies offer AIGeschikt white-label through the Partner package: €899 per month excluding VAT for 3 clients, plus €149 per additional client. Your client sees your brand: reports, dashboard and communication run under your name. You keep the client relationship and set your own price; AIGeschikt measures, writes and publishes. You approve proposals on your client's behalf or let your client do it, and you receive one invoice.

Also covered in Pricing · AIGeschikt vs an SEO agency

An agency connects each client site to AIGeschikt separately. On WordPress you install AIGeschikt's connector plugin on your client's site; other CMSs connect through their own integration. After that the measurement runs on its own: 1,200 measurements every week across that client's buying questions, put to ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews. There is no implementation project, so you start with one client and add the rest whenever you want.

AIGeschikt is a brand name of VSMD V.O.F., located at Westzeedijk 102 (unit 1), 3016 AH in Rotterdam. The business is registered with the Dutch Chamber of Commerce (KVK) under number 42097283, with VAT identification number NL869705234B01. You can reach AIGeschikt at [email protected], in Dutch or English. The service is built for the Dutch market.

Also covered in Who is behind AIGeschikt

AIGeschikt was founded by Maurits Dekker and Vincent Schrooyen, both recent graduates of Erasmus University Rotterdam. During his bachelor's at TU Delft, Maurits Dekker started his own web design and marketing agency; at AIGeschikt he builds the technology and new features. Vincent Schrooyen worked at Accenture for half a year and leads strategy and delivery to clients. You are in direct contact with the people who build your AI visibility.

Also covered in The team