Getting found in ChatGPT starts with access: allow OAI-SearchBot and make sure search engines know your site. After that, what counts is pages that answer one question clearly, with figures and sources, and mentions on other sites. Measure it monthly with a fixed list of questions, because there are no guarantees.
That is why anyone looking for an SEO agency today hears the term GEO more and more: generative engine optimization, making sure AI answers name your brand. Below is what you can do yourself, in the order in which it counts.
Why ChatGPT first? In Belgium in August 2026, 78.17% of the traffic that chatbots send to websites came from ChatGPT, compared with 11.13% from Gemini and 2.21% from Perplexity [1], [2]. What follows works just as well for the others.
Step 1: let the right bots in
OpenAI uses several crawlers, and each one does something different [3]:
- OAI-SearchBot shows websites in ChatGPT's search results. Sites that block it do not appear there, except at most as a navigation link.
- GPTBot collects content to train future models. Your visibility in search does not depend on it.
- ChatGPT-User fetches a page when a user asks for it; robots.txt may not apply there.
OpenAI picks up a change to your robots.txt within about 24 hours [3]. PerplexityBot follows the same logic: it shows and links sites and does not train models [4]. This robots.txt allows the search functions of ChatGPT and Perplexity, but refuses training by OpenAI:
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: GPTBot
Disallow: /
Allowing or refusing training is a business decision, not a technical one. Do be careful with Google-Extended: blocking that token also refuses grounding in the Gemini apps, although it changes nothing in Google Search [5].
Look beyond your robots.txt. If your site sits behind Cloudflare, bots can be stopped at the gate while your robots.txt allows everything. Since 1 July 2025, every new domain chooses at sign-up whether AI crawlers are allowed in [6], and Cloudflare announced that from 15 September 2026, training and agents will be blocked by default for new domains on pages with ads [7].
If you block "training" in Cloudflare, you also block crawlers that combine search and training, such as Googlebot and Bingbot [7]. So check your dashboard, or ask whoever manages your hosting.
Put your content in text. Prices, services and contact details belong in the HTML as plain text, not only in an image, a PDF or a block that only appears after loading. Google recommends this [8]; for the OpenAI and Perplexity crawlers it is not documented whether they execute JavaScript.
Step 2: make sure search engines know you
ChatGPT search uses external search engines, plus content that partners supply directly [9]. Google AI Overviews and AI Mode draw on the regular Google index, with no extra requirements: a page must be indexed and eligible to be shown with a snippet [8]. So AI visibility starts with ordinary indexing.
- Register your site in Google Search Console and Bing Webmaster Tools, and submit your sitemap. OpenAI does not say which search engines it uses; adding Bing takes ten minutes.
- Check that your important pages do not carry
noindexornosnippet. Withnosnippetandmax-snippetyou limit what Google shows in AI answers [8].
Not sure whether that is in order? The free scan looks at indexing and also makes a baseline measurement of your visibility in ChatGPT, Gemini, Perplexity and Google AI Overviews.
Step 3: write answers that can be cited
Researchers from Princeton and elsewhere, who introduced the term GEO, modified source texts and measured how visible they then were in generated answers [10]. Citing sources, adding quotations and adding statistics gave 30 to 40% more visibility; keyword stuffing yielded little or nothing. Smaller players gained the most: a source in fifth place became 115% more visible with citations [10].
That was a test set-up with models from 2023, not a measurement in today's ChatGPT. But the direction matches what Google recommends, and it is exactly what a reader wants:
- One page, one question. "How much does a roof renovation cost per square metre?" deserves its own page.
- The answer in the first two sentences, then the nuance.
- Figures with a source and date, and a visible date of the last update.
- Who, what and where in plain sentences. "Schrijnwerkerij Peeters installs wooden windows in the Leuven region" (a made-up example) says more than "we take care of everything, with passion".
- An FAQ with questions customers really ask, in plain text.
How AI systems choose between sources is covered in how ChatGPT, Gemini and Perplexity choose their sources.
Step 4: become a brand others write about
The same details everywhere. Your name, address and activity should be identical on your website, your Google Business Profile, LinkedIn, review platforms and in the KBO, the Belgian company register. An Organization block in structured data makes those facts machine-readable, but Google says you need no special markup for AI features, and that markup must match what is visible on the page [8]. So it is hygiene, not a lever; and you don't need an llms.txt file for Google either [8].
Mentions elsewhere. Ahrefs studied 75,000 brands. The number of times a brand was mentioned on the web correlated most strongly with visibility in Google AI Overviews (correlation 0.664), much more strongly than the number of backlinks (0.218) [11]. Correlation does not prove causation, and the study covered AI Overviews, not ChatGPT [11].
The lesson stands: make sure you are mentioned where your customers read, such as your trade federation, the regional press, partner pages and reviews that name the problem you solved. This is the slowest step, and nobody trusts paid "top 10" lists on unknown sites, neither people nor models.
Step 5: measure with a fixed list of questions
- Draw up 20 to 30 questions your customers really ask, in Dutch and French if you sell in both languages.
- Ask each question several times in each engine, because answers vary. Note whether you are mentioned, alongside which competitors and with which sources.
- Repeat this monthly with the same list. A trend over three months says more than one test.
- Track the traffic. ChatGPT adds
utm_source=chatgpt.comto links to your site [12], so you can see those visits separately in Google Analytics. Cédric explains how to read UTM codes in UTM codes explained.
You can do this by hand in a spreadsheet. A tracker does it automatically: within ClickTrust Group we built our own tool for this, aiviz, from €50 a month.
What GEO cannot do
- Give guarantees. Nobody controls what ChatGPT answers. Anyone who promises you "position one in ChatGPT" is selling something they have no control over.
- Be bought. Since 24 August 2026 ChatGPT does show ads in Belgium, as a sponsored card below the answer. How that works and what it costs is covered in ChatGPT Ads in Belgium.
- Change the model's memory quickly. What a model knows about you from its training only changes with the next model generation. Live search moves faster.
- Work without SEO. Indexing, fast pages and clear content are the basis of both.
Building a new site soon? Then include this in the briefing instead of fixing it afterwards. What that costs is covered in how much does a website cost in 2026?
Frequently asked questions
Can I pay to be named in ChatGPT's answer?
No. Since 24 August 2026 ChatGPT does show ads in Belgium, but they appear as a sponsored card below the answer and change nothing about what ChatGPT itself says. You earn organic mentions with access, clear content and mentions elsewhere.
Do I need to allow GPTBot to appear in ChatGPT?
Not for search. OpenAI uses OAI-SearchBot to show sites in ChatGPT search; GPTBot collects content to train future models. You can allow OAI-SearchBot and refuse GPTBot, or the other way round.
Do I need an llms.txt file?
Not for Google: Google writes that you need no new machine-readable files or AI text files to appear in AI Overviews or AI Mode. A clean HTML page with the right bots allowed does more than an extra file.
How quickly will I see results from GEO?
According to OpenAI and Perplexity, an updated robots.txt is picked up within about 24 hours. New pages count as soon as they are crawled and indexed. What a model knows about you from its training only changes with the next model generation. There are no guarantees.
What does GEO cost?
That depends on your starting point. Sorting out access is a few hours' work; building content and mentions is an ongoing programme. At Izoard, GEO is part of the SEO programme, tailored after a free scan. Our tracker aiviz, which follows your AI visibility, costs from €50 a month.
Sources
- AI Chatbot Market Share Belgium, August 2026 — Statcounter Global Stats, accessed 27 Sep 2026
- New Statcounter AI data finds ChatGPT sends 79.8% of all chatbot referrals to websites — Statcounter Global Stats, accessed 27 Sep 2026
- Overview of OpenAI Crawlers — OpenAI, accessed 27 Sep 2026
- Perplexity Crawlers — Perplexity, accessed 27 Sep 2026
- Google's common crawlers (Google-Extended) — Google Search Central, 14 Jul 2026, accessed 27 Sep 2026
- Cloudflare Just Changed How AI Crawlers Scrape the Internet-at-Large — Cloudflare, 1 Jul 2025, accessed 27 Sep 2026
- Your site, your rules: new AI traffic options for all customers — Cloudflare Blog, 1 Jul 2026, accessed 27 Sep 2026
- AI features and your website — Google Search Central, 10 Dec 2025, accessed 27 Sep 2026
- Introducing ChatGPT search — OpenAI, 31 Oct 2024, accessed 27 Sep 2026
- GEO: Generative Engine Optimization (KDD 2024) — Aggarwal et al., arXiv 2311.09735, 28 Jun 2024, accessed 27 Sep 2026
- An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) — Ahrefs (Louise Linehan), 26 May 2025, accessed 27 Sep 2026
- Publishers and Developers - FAQ — OpenAI Help Center, accessed 27 Sep 2026