How AI Search Engines Choose Which Pages to Cite
If you want to get cited in AI Overviews, it helps to know how these tools work. Google AI Overviews, AI Mode, ChatGPT search and Perplexity all work in a similar way. They search an index, pull back pages that match, and write an answer that links to some of them. Your goal is to be one of the pages they find, and then the one they choose to quote.
Google calls the main method query fan-out. The system breaks your question into several related smaller searches, runs them all, and builds one answer from what comes back. That means you can be cited for a sub-question even when you do not rank for the main search.
- Retrieval: the system finds possible pages in its search index.
- Grounding: the model bases its answer on those pages, not only on what it learned in training.
- Citation: it links to the pages that back up specific claims in the answer.
So in my view, “AI visibility” mostly comes down to two things you already know. Get indexed, and have the clearest, most useful passage on the subject.
What Google Officially Says About AI Overviews and AI Mode
Google’s position is blunt. There are no extra requirements to appear in AI Overviews or AI Mode. If your page is indexed and eligible to show with a snippet, it can be used.
Google’s AI features and your website page (last updated December 2025) lists ordinary SEO basics as the way in. In May 2026, Google added a longer guide to optimizing for generative AI features. It says AEO and GEO for Google are simply SEO. I agree, and I find that a relief for most site owners.
What Google Recommends
- Allow crawling in robots.txt and in any CDN or firewall rules.
- Link to important pages from your other pages, so they are easy to find.
- Keep key content in text, not locked inside images or scripts.
- Write unique content based on real experience, not generic “tips” pages.
- Use clear headings, sections and paragraphs.
- Add good images and video where they help.
- Keep structured data in line with what is visible on the page.
- Keep Merchant Center feeds and your Business Profile up to date.
Google also warns against mass-producing pages to game AI answers. That can fall under its scaled content abuse policy. The effects can show up during updates like the ones in our Google core updates 2026 tracker.
Why AI Overviews and AI Mode Can Cite Different Pages
Google says the two features may use different models and methods. AI Overviews only appear when Google thinks they add something to normal results. AI Mode handles longer, comparison-style questions and often shows a wider spread of links. Our roundup of Google I/O 2026 Search changes covers how the two features connect.
How ChatGPT Search and Perplexity Pick Sources
Neither OpenAI nor Perplexity publishes ranking factors. What they do document is crawler access. If their search bot cannot reach your page, you will not be cited in their answers. Our guide to how Perplexity chooses its sources covers what is known.
ChatGPT Search
OpenAI’s crawler documentation says you must allow OAI-SearchBot to appear in ChatGPT search answers. Sites that block it can still appear as plain links for people looking for that site, but not as cited sources.
- OpenAI says it also uses third-party search providers, but it has not named them officially.
- Many SEOs make sure their pages are indexed in Bing as a sensible safety step. I do this too. But treat it as common practice, not confirmed policy.
- robots.txt changes for search can take around 24 hours to take effect.
Perplexity
Perplexity’s crawler docs say PerplexityBot finds and links sites in its search results. Perplexity says this bot is not used to collect training data for foundation models.
Both platforms also have agents that a user triggers (ChatGPT-User and Perplexity-User). These fetch a page when someone asks about it, and robots.txt rules may not apply to them.
Make Sure Every AI Search Bot Can Crawl You
Crawl access is the one hard requirement on every platform. Check robots.txt, your CDN bot settings and your firewall. A security rule can quietly block a bot that robots.txt allows. When I audit a site for AI visibility, this is always where I start.
| Bot | Used for | Block it and you lose |
|---|---|---|
| Googlebot | Google Search, including AI Overviews and AI Mode | All Google Search visibility |
| Google-Extended | Training and grounding in Gemini and other Google AI products | Nothing in Google Search; it doesn’t control AI Overviews |
| OAI-SearchBot | ChatGPT search results | Citations in ChatGPT search answers |
| GPTBot | Training OpenAI models | Nothing in ChatGPT search directly |
| PerplexityBot | Perplexity search results | Citations in Perplexity answers |
| Bingbot | Bing and Copilot | Bing, Copilot and partner surfaces |
A common mistake I see is blocking every “AI bot” at once. If you want to stay out of model training but still be cited, block the training crawlers and allow the search crawlers.
Quick Crawl Checks
- Open yoursite.com/robots.txt and search for each user agent in the table.
- Check your CDN or WAF (web application firewall) for “block AI bots” switches.
- Run key URLs through the URL Inspection tool in Search Console.
- Confirm the main content appears in the rendered HTML, not only after someone clicks or scrolls.
For a full process, use our technical SEO audit checklist.
Write Answer-First Pages That Are Easy to Quote
Put the direct answer near the top of each section, then explain it. AI systems quote passages. A clear answer that stands on its own under a matching heading is far easier to cite than a point buried in paragraph six.
This is not a special AI trick. Google says you do not need to “chunk” pages or rewrite them for machines. It is simply good writing, and it happens to suit how passages get pulled into answers. It is how I’d write the page even if AI search did not exist.
How to Structure a Section That Gets Cited
- Use a heading that matches a real question, such as “How long does a roof replacement take?”
- Answer it in the first one or two sentences, with a specific number or fact.
- Add the detail: conditions, exceptions, steps or examples.
- Use a list or table when you compare options or give steps.
Add Something AI Cannot Make on Its Own
Google’s guide says content a model could easily write adds little value. Give the systems a reason to cite you instead of summing up everyone else:
- Original data, tests, prices or results from your own work.
- First-hand photos, screenshots and worked examples.
- Clear expert opinions, with the person’s name and qualifications.
- Specific local or industry detail that generic pages miss.
Make Your Entities and Facts Clear
AI systems need to know exactly who you are and what you are talking about. An “entity” here just means a clearly defined thing, like your brand, a product or a person. Use the same names every time, state key facts plainly, and make the same details match across your site and the wider web.
- Use your brand, product and people names the same way everywhere.
- Name things specifically: “Google Search Console”, not “the tool”.
- Add author bylines and bios that show real experience.
- Keep an About page with your business name, location and what you do.
- Make sure your Google Business Profile matches your site for address, hours and services.
Mentions on other trusted sites help AI systems link your brand to a topic. Earn them the normal way, as covered in our guide to link building that works. Google warns clearly that fake mentions on blogs, videos or forums will not help. For how mentions compare with links, see brand mentions vs backlinks.
Use Schema for What It Really Does
Structured data is not required for AI features, and there is no special “AI schema”. Keep using standard schema.org markup for rich results and to describe your content clearly.
Google says this plainly in its AI optimization guide. Schema can still help in normal ways:
- Organization and Person markup make clear who is behind the content.
- Product, Review and LocalBusiness markup support rich results and shopping or local features.
- Article markup with accurate dates helps describe the page.
The rule that matters most: markup must match what is visible on the page. Marking up facts users cannot see is a policy problem, not an improvement.
Keep Content Fresh and Accurate
AI answers are based on up-to-date pages from the index, so old facts work against you. Update pages when the facts change, not on a fixed schedule just to change the date.
- Review pages with prices, statistics, product specs or rules at least every three months.
- Show a visible “last updated” date when you make a real change.
- Remove or redirect old pages that contradict newer ones.
- Cut down duplicate pages, so systems do not pick an old version.
- Keep your XML sitemap’s lastmod values honest, so crawlers find changes.
Changing a date without changing the content does not help. It can also hurt trust if readers notice. I’d skip this tactic completely.
Control What AI Features Can Show
You have several controls, from limits on single pages to an opt-out for the whole site. Use them on purpose, because each one trades visibility for control.
Page-Level Controls in Google
- nosnippet: stops Google showing a text snippet, which also keeps the page out of AI features.
- data-nosnippet: hides one section of a page from snippets and AI answers.
- max-snippet: limits how much text Google can show.
- noindex: removes the page from Search completely.
Changes only apply after Googlebot crawls the page again. Google notes that can take days or months, so use URL Inspection to request a new crawl.
The Search Generative AI Control
Search Console now has a setting for the whole site that removes your content from AI Overviews, AI Mode and AI Overviews in Discover. Google announced it on 3 June 2026. It says the setting rolled out to all sites worldwide by 31 August 2026.
Google says opting out is not used as a ranking signal for normal results. But you lose all impressions and clicks from those AI features. So most sites that want citations should leave it alone. That is what I recommend unless you have a strong business reason.
Other Platforms
- ChatGPT: block OAI-SearchBot to leave its search answers; use noindex to be removed fully.
- Perplexity: block PerplexityBot in robots.txt.
How to Measure AI Overviews, AI Mode and ChatGPT Traffic
Use Search Console for Google’s AI features, and GA4 for visits from chatbots. Neither gives you the full picture, so I treat the numbers as a minimum, not the total.
Search Console
Clicks and impressions from AI Overviews and AI Mode are counted in the normal Performance report, under the Web search type. Since mid-2026 there is also a Generative AI performance report for Search, plus a separate one for Discover. Our guide explains how to read the Generative AI reports.
- It shows impressions only, with no click or query data at the time of writing.
- You can break results down by page, country, device and date.
- It only appears once your site has enough AI impressions.
GA4
Clicks from AI Overviews and AI Mode arrive as google / organic, so you cannot separate them in GA4. Visits from chatbots are easier to spot.
- GA4 added an “AI Assistant” channel to its default channel group in May 2026. Google names ChatGPT and Gemini as examples, but has not published a full list of sources.
- ChatGPT adds utm_source=chatgpt.com to many links, which helps you see where visits came from.
- Build a custom channel group with a source regex (chatgpt, perplexity, copilot, gemini, claude) to catch visits the default channel misses.
- Expect some AI visits to show as Direct. Apps and in-app browsers often drop the referrer (the information about where a visit came from).
Bing Webmaster Tools
Bing’s AI Performance report has been in public preview since early 2026. It shows citations in Copilot and Bing AI answers, including which pages were cited and sample grounding queries.
Myths About Getting Cited by AI
Most “AI SEO” myths sell a shortcut that does not exist. Here is where the main claims stand right now.
| Claim | Status |
|---|---|
| You need llms.txt to appear in AI answers | False for Google, which says Search ignores it. OpenAI’s and Perplexity’s crawler docs don’t mention it either. |
| There’s special AI schema | False. No schema.org type exists for AI features. |
| You must split pages into small chunks | False. Google says its systems find the relevant section of a long page. |
| You should rewrite content in an “AI style” | False. Write for people; systems understand synonyms and intent. |
| Blocking GPTBot removes you from ChatGPT search | False. ChatGPT search uses OAI-SearchBot. |
| Google-Extended controls AI Overviews | False. AI Overviews are controlled by Googlebot and the Search Console setting. |
| A tool can show Google’s AI ranking data | False. Google says no third-party tool can access its ranking or AI systems. |
An llms.txt file will not hurt, and some tools may read it. Just do not expect it to earn citations. And please do not pay anyone who promises guaranteed AI placement. Our guide to LLMs.txt explains what the file does.
Next Steps
Here is the order I would work in if you want to get cited in AI Overviews and other AI answers:
- Check robots.txt, CDN and firewall rules so Googlebot, Bingbot, OAI-SearchBot and PerplexityBot can all crawl you.
- Check the Search generative AI setting in Search Console is set the way you want.
- Pick your ten most important pages and rewrite each section to answer first.
- Add original data, examples or first-hand detail to each of those pages.
- Tidy up author bios, your About page and your Business Profile so your details match.
- Set up a GA4 custom channel group for AI visits, and review the Generative AI report every month.



