How to Optimize Your Website for AI Search Engines
OpenAI launched ChatGPT search on October 31, 2024, and since February 5, 2025 anyone can use it without an account. Google has shown AI Overviews in Dutch since May 20, 2025 and AI Mode since October 8, 2025. On October 8, 2025, Microsoft wrote that Copilot runs on the Bing search index. Google's documentation on AI features (updated December 10, 2025) says AI Overviews and AI Mode have no extra technical requirements and need no separate optimization. A few technical points are still worth checking, starting with which crawlers your site lets in.
The short version
Optimizing a website for AI search engines mostly means getting the basics right: AI crawlers can read your site, your most important answers sit on the page as plain text, and visitors can see who is behind the content. Google says its AI Overviews and AI Mode need no separate optimization. For ChatGPT, Perplexity and Copilot, the main job is making sure their own crawlers are not blocked.
What should you check on your website?
Work through the nine items below in order. Access comes first, then content, then extras. The table shows what to check for each item.
| Item | What to check | Where to look |
|---|---|---|
| Crawlers | Whether search crawlers from Google, Bing, OpenAI, Perplexity and Anthropic are allowed in | robots.txt, firewall, CDN |
| Conversational content | Whether a short answer sits under every heading | Your most important pages |
| Structured data | Whether the markup matches the visible text | Rich Results Test, Search Console |
| E-E-A-T | Whether it is clear who wrote the page and why they know the subject | Author line, about page |
| Technical SEO | Status code 200, indexable, fast, text as text | Search Console, page speed |
| Visual content | File names and alt text | Your media library |
| Local visibility | Identical business details everywhere | Google Business Profile, Bing Places, your site |
| llms.txt | Whether a file Google ignores is worth your time | Root folder of your site |
| AI agents | Whether buttons, links and forms work for an agent | HTML and accessibility |
1. Let in the AI crawlers you want
Providers run separate crawlers for search, for training models and for user requests. You control them in robots.txt, a text file in the root folder of your site that lists which bots may fetch which pages. As of October 6, 2026, the providers' documentation says the following, including OpenAI, Perplexity and Anthropic:
- OpenAI: OAI-SearchBot fetches pages for ChatGPT's search results, and a site that blocks it does not appear in ChatGPT search answers. GPTBot collects content that may be used to train models. ChatGPT-User visits pages when a user asks it to, and according to OpenAI, robots.txt rules do not always apply to it. The settings are independent of each other. A change to robots.txt can take about 24 hours to take effect for search results.
- Google: AI Overviews and AI Mode use the regular search index, so Googlebot has to be able to reach your pages. The Google-Extended token controls whether your content is used to train Gemini models and for grounding in the Gemini apps, where answers are supplemented with content from the search index. According to Google, it has no effect on Google Search. If you want to keep pages out of AI answers instead, use nosnippet, data-nosnippet, max-snippet or noindex.
- Perplexity: PerplexityBot fetches pages for Perplexity's search results, and Perplexity says it is not used to train models. Perplexity-User fetches pages on behalf of a user and generally ignores robots.txt.
- Anthropic: ClaudeBot collects content for training, Claude-SearchBot indexes content for search, and Claude-User fetches pages when a user asks a question. You block them through robots.txt. Anthropic also publishes IP addresses but says blocking by IP address does not always work well.
- Microsoft: Copilot runs on the Bing search index, so Bingbot must not be blocked in your robots.txt.
Look at your hosting and security setup too. Google lists crawling that robots.txt, your CDN and your hosting infrastructure all allow as a basic requirement. OpenAI and Perplexity recommend allowing their published IP addresses in your firewall. Since July 1, 2025, Cloudflare asks for every new domain whether AI crawlers get access, so check that setting if you use Cloudflare. Blocking GPTBot does not remove you from ChatGPT search, because OpenAI says the settings work independently.
How can you check in five minutes whether your site is open?
Open yourdomain.com/robots.txt in your browser and look for Disallow rules next to the bot names in the list above. A rule like "User-agent: *" followed by "Disallow: /" shuts out every bot, search bots included. Then open the admin screen of your hosting or CDN and see whether a bot filter is active that rejects traffic from unknown crawlers. Many web shops and WordPress sites have one switched on without the owner knowing.
2. Write conversational content
On May 21, 2025, Google wrote that people ask longer and more specific questions in AI Overviews and AI Mode, with follow-up questions on top. For Copilot, Microsoft recommends headings that name the topic, question-and-answer blocks, lists and tables. It also warns against putting important answers in tabs or expandable menus, since AI systems may not render hidden content. Google says you do not need to write specially for AI, split your pages into small pieces or cover every long-tail variation, because AI systems understand synonyms.
Keep it practical: put a short answer under each heading first, then expand on it in language your customer understands. You can add frequently asked questions to a page as a question-and-answer block. What happened to the matching markup is covered below.
What does a page that is easy to cite look like?
An easily citable page opens with a short paragraph of two to four sentences that answers the main question. Below it, headings name a question or topic, and every paragraph makes sense when someone reads it on its own. That means naming the subject instead of pointing to a sentence elsewhere with "this" or "they". An AI system often lifts a single paragraph from your page, and that paragraph has to hold up by itself.
Take an installation company writing about whether a heat pump suits a house built in the 1970s. Under the heading, a two-sentence answer comes first, with the conditions that apply. Insulation, heat emission system and costs follow. A customer finds that pleasant to read, and it is also a shape an AI system can easily pick up.
3. Use structured data (schema markup)
Structured data is code on your page, usually JSON-LD, that tells search engines in a fixed format what kind of page it is: an article, a business, a product or an event. Google says structured data is not required for AI Overviews and AI Mode, and that no special schema.org markup exists for them. The markup does have to match what is visible on the page. On October 8, 2025, Microsoft wrote that schema, usually as JSON-LD, helps search engines and AI systems understand your content. So markup can pay off for Bing and Copilot. Google's AI features do not require it, but it stays useful for rich results.
Not every type recommended in 2024 still works. Google stopped showing FAQ rich results on May 7, 2026 and removed the documentation in June 2026, according to its Search documentation changelog. It stopped showing how-to results on September 13, 2023. Google said in 2023 that you do not have to remove existing FAQ markup, but it no longer earns you anything in Google. Supported types include Organization, LocalBusiness, Article, Breadcrumb, Product, Review snippet and Event.
For a small business site, Organization or LocalBusiness on the homepage and Article on blog posts are the most logical choices. Test your markup with Google's Rich Results Test, and after a few weeks check Search Console for reported errors.
4. Work on E-E-A-T
E-E-A-T stands for Experience, Expertise, Authoritativeness and Trustworthiness. On December 15, 2022, Google added Experience to E-A-T in the guidelines for its quality raters, and the concept has been called E-E-A-T since. According to Google, trustworthiness is the most important part, and E-E-A-T itself is not a ranking factor. Google's systems do use a mix of signals that identify good E-E-A-T, and they weigh it more heavily for topics that affect health, financial stability or safety. Google asks you to show who made your content, how it was made and why it exists. So put a name on your articles, link it to a page about the writer, and explain how a piece came about, even if AI helped write it. Invented author profiles with AI-generated faces or fake titles make a page untrustworthy.
In practice that means a real name and a short profile line on every article, an about page introducing the people behind the company, contact details that are correct, and examples from your own work. The examples are the strongest of these for a small business, because a big competitor or a language model cannot fake your experience with your customers.
5. Technical SEO and speed
Google counts a good page experience among the things that matter for AI features: the page works on every device, loads fast and has a main content area that is easy to identify. The page has to return a successful status code (HTTP 200) and have indexable content. Important information belongs on the page as text. Microsoft advises against putting key information only in PDFs or images. If your site relies heavily on JavaScript, follow Google's JavaScript SEO guidelines.
Internal links remain the way to make your content discoverable, Google says. Microsoft still names crawlability, metadata, internal links and backlinks as the basis of visibility. A technically sound site, such as a fast website with a logical structure, helps search engines and AI systems alike.
6. Give visual content context
Google writes that people can take or upload a photo with AI and ask a question about it. Support your text with good-quality photos and videos. Give images a descriptive file name and alt text: Google uses alt text to understand what an image is about. Microsoft advises against putting important information only in an image, because that makes it harder and less accurate for AI systems to read.
On March 26, 2026, Google began rolling out Search Live worldwide in every language and region where AI Mode is available: people talk to Search and can use their camera. Since September 24, 2026, Search Console also reports on searches with an image in a separate filter, including Lens, Circle to Search and uploaded images.
Good alt text describes what you can see, for example "technician installing a heat pump on the facade of a row house". "IMG_4521.jpg" and "image" tell a search engine nothing.
7. Look after your local visibility
According to Google, AI answers can include information about local businesses, and an up-to-date Google Business Profile (formerly Google My Business) helps your services show up in AI answers and in regular search results. The profile is free and lets you manage how your business appears in Google Search and Maps. Microsoft has Bing Places for Business for this, which was redesigned in October 2025. Make sure your name, address, phone number, opening hours and website match on both platforms and on your own site.
8. llms.txt and what providers do with it
llms.txt is a proposed file in the root folder of your site, meant as a summary for AI systems. Google's guide to AI features (May 15, 2026) says Google Search ignores the file and that it neither helps nor hurts your visibility there. In June 2025, Google's John Mueller wrote on Bluesky that no AI system used it at that point. In Ahrefs' study of 137,210 websites, 28 percent publish the file, but in May 2026, 97 percent of those files received no requests at all. AI search bots made 1.1 percent of all requests, with coding agents making more. The crawler documentation from OpenAI, Anthropic and Perplexity does not mention the file. According to Ahrefs, you do not need it to appear in AI answers.
Creating the file takes little time, but the payoff has not been shown so far. Spend your time on the items above instead, since they do change how people and search engines read your pages.
9. Plan for AI agents
Google describes how browser agents can view a site through screenshots, the HTML structure and the accessibility tree to carry out tasks such as making a reservation. The explainer on web.dev (updated April 1, 2026) gives practical pointers: use real button and link elements instead of custom divs, associate labels with input fields, do not lay invisible layers over buttons, and keep the layout stable. That is plain accessibility work too. WebMCP, a proposed web standard that lets a site offer agents explicit tasks, is still in development, according to web.dev.
What can you do today?
- Open your robots.txt and check that OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot and Bingbot are not blocked.
- Check the AI crawler setting in your CDN or hosting (Cloudflare, for example).
- Pick your five most important pages and put an answer of two to four sentences at the top of each.
- Add an author name and a short profile line to your blog posts.
- Compare name, address, phone number and opening hours in Google Business Profile, Bing Places and on your site.
- Give your most important images descriptive alt text.
If you want someone to look into this for your own site, a specialist in AI and search visibility can review your robots.txt, markup and copy.
Frequently asked questions
Do I need to change my website for ChatGPT and Google AI?
Google says AI Overviews and AI Mode need no separate changes. For ChatGPT search, OAI-SearchBot must be able to read your site. Start with your robots.txt and firewall, and make sure your most important answers appear on the page as plain text.
Should I block or allow AI crawlers?
That depends on what you want. If you want to be mentioned in ChatGPT, Perplexity or Claude, allow their search crawlers. Training and search are controlled separately, so you can refuse training bots and allow search bots. If you block everything, your site probably will not show up in those answers.
Do I need an llms.txt file?
No. According to Google's guide of May 15, 2026, Google Search ignores the file, and Ahrefs' study shows it was almost never requested in May 2026. It does no harm, but right now you get no demonstrable visibility from it.
Is schema markup still worth it?
Yes, with limits. Google turned off FAQ rich results on May 7, 2026. Types such as Organization, LocalBusiness, Article and Product are still supported and help search engines understand your page. The markup always has to match what you show on the page.
What is E-E-A-T and why does it matter?
E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness. It comes from the guidelines Google uses for its quality raters. Trustworthiness is the most important part. You show it with a real author, an explanation of how your content was made, and examples from your own work.
What Make it Matter does for you
Make it Matter helps businesses prepare their websites for AI search engines, with content creation, technical SEO and schema markup.
