AI search & content ops

How to Get Cited by ChatGPT, Perplexity and Google's AI Overviews

What each AI search product documents about how it finds and cites sources, the crawler access each one needs, what to change on the page, what not to bother with, and how to measure AI referrals with the data that exists today.

Ulrich Svarrer

November 15, 20268 min read

AI search products cite pages they can reach through their own search index or crawler, and that answer the question well. There is no separate trick. To be cited by ChatGPT, Perplexity, Microsoft Copilot and Google’s AI Overviews and AI Mode, a page needs three things: the right crawler access for each product, eligibility in each product’s index, and content specific enough to be worth citing. This guide covers what each product actually documents, a robots.txt that allows search while blocking training if you want that, what to change on the page, what not to bother with, and how to measure AI referrals and citations with the data that exists today.

AI search is moving quickly, so every claim here links to the product’s own documentation, and this guide will be reviewed as that documentation changes. For the broader question of where AI helps a content team, see AI for content operations.

How each product finds and cites sources

Product	Where sources come from	Access it needs	Where to measure
Google AI Overviews and AI Mode	Google’s Search index	Googlebot; the page indexed and eligible for a snippet; the Search generative AI setting left on	Search Console: Performance (Web) and the generative AI report
ChatGPT search	OpenAI’s search crawler and index	OAI-SearchBot allowed in robots.txt	GA4: the AI Assistant channel, or referrals tagged utm_source=chatgpt.com
Perplexity	Perplexity’s crawler and index	PerplexityBot allowed in robots.txt	GA4: referrals from perplexity.ai
Microsoft Copilot and Bing	Bing’s index	Bingbot; the page indexed in Bing	Bing Webmaster Tools: AI Performance; GA4 referrals

The common thread is that each product cites from an index it builds with its own crawler. If that crawler cannot reach a page, nothing on the page matters. That makes access the first thing to check.

Step 1: Check crawler access

Most AI companies run separate crawlers for search and for training models, and you can treat them differently:

OpenAI: OAI-SearchBot is used to surface sites in ChatGPT search; OpenAI says sites that block it will not be shown in ChatGPT search answers. GPTBot collects training data and is set independently. ChatGPT-User fetches a page when a user asks, and robots.txt may not apply to it. See OpenAI’s crawler documentation.
Perplexity: PerplexityBot surfaces sites in Perplexity’s results and, according to Perplexity, is not used to train foundation models. Perplexity-User fetches pages for user requests and generally ignores robots.txt.
Google: AI Overviews and AI Mode use the normal Search crawler. Google-Extended is a separate control for training and grounding in Gemini apps and Vertex AI; Google says it does not affect inclusion in Google Search.
Microsoft: Copilot draws on Bing. Since October 2026 Bing also offers separate opt-outs for training (in robots.txt) and for grounding (per page, in a meta tag); both default to allowed.

A robots.txt that allows AI search crawlers and blocks training crawlers looks like this. Adapt it to your own decision; there is no right answer for every business:

# Allow AI search, so pages can be cited
User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

# Block model training
User-agent: GPTBot
Disallow: /

User-agent: Google-Extended
Disallow: /

Then check what actually reaches your server. Firewalls and CDN bot settings can block crawlers that robots.txt allows; some block AI crawlers by default. Look for each user agent in your server or CDN logs, and verify requests against the IP ranges the companies publish. OpenAI says robots.txt changes take about 24 hours to apply.

Step 2: Be eligible in Google

Google’s AI features documentation is short and clear: there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible to show with a snippet, and no special markup is needed. Two controls matter:

Snippet controls. nosnippet, data-nosnippet and max-snippet limit what Google can show from a page, in AI features as well as regular results. Use them where you need to, knowing they reduce what can be cited.
The Search generative AI setting in Search Console lets you include or exclude a site from AI Overviews, AI Mode and generative features in Discover. Sites are included by default. Check nobody switched it off.

Step 3: Make the page worth citing

No AI search product publishes a formula for which pages it cites. What Google does say, in its 2025 advice on AI experiences and its 2026 optimization guide, is that optimizing for AI features is still SEO, and that unique, non-commodity content is what does well. The practical version, for a content team:

Do	Why it fits how these products work
Answer the question plainly, early on the page	A summary needs a clear statement to rely on; readers who click through want it too
Publish what a summary cannot produce: original data, worked examples, tested results, current figures	A generic explanation is already in the model; a specific fact needs a source
Make facts checkable: sources, dates, named authors	The same trust signals Google’s guidance asks for in all content
Keep facts current, with an honest updated date	Outdated figures are a reason not to cite a page; see what counts as a real content update
Use clear headings and tables where they fit the content	They help readers and any system extracting a specific answer
Keep structured data accurate and matching the visible page	Google asks for this; it is not a requirement for AI features

A useful test: if someone asked a knowledgeable colleague this question, would your page be the one they forwarded? If the page only restates what every other result says, a summary can say it too, without the link. Our helpful content rubric is a practical way to check.

What not to bother with

Claim	What the documentation says
“Add an llms.txt file to appear in AI answers”	A proposal, not a standard. Google’s optimization guide says Google Search ignores llms.txt files. We found no statement from OpenAI or Perplexity that they use them
“Chunk and rewrite content for AI”	Google’s guide lists chunking and rewriting content for AI among the things you do not need to do
“Schema markup is required for AI Overviews”	Google says no special markup is needed
“Create pages for every question the AI might search”	Google warns that content made mainly to target the extra searches its AI features run can violate its scaled content abuse policy
“Seed brand mentions across the web”	Google’s guide names inauthentic mentions as something not to do
“GEO raises visibility by 40%”	That figure comes from a research benchmark built for the 2023 GEO paper, not from measurements of ChatGPT or Google
“Block GPTBot and you will vanish from ChatGPT”	GPTBot is for training; ChatGPT search uses OAI-SearchBot

Be sceptical of citation statistics in vendor articles, too. Most come without a method, and answers vary by user, location and time, which makes general percentages hard to trust.

Measuring AI citations and traffic

Measurement has improved a lot since 2025. Use the sources the products provide, and be clear about what each one counts:

Source	Counts	Does not count
Search Console, generative AI report	Impressions in AI Overviews, AI Mode and Discover’s AI features, by page, country, device and date	Clicks or queries; those stay in the Performance report under Web
Bing Webmaster Tools, AI Performance (preview since February 2026)	Citations in Copilot and Bing’s AI answers, cited pages and a sample of grounding queries	Clicks
GA4, AI Assistant channel	Sessions referred by AI assistants such as ChatGPT, Gemini and Copilot, in the default channel group	AI Overviews and AI Mode, which count as Organic Search
A fixed prompt sample	Whether your pages are cited for 20 to 30 questions you care about, checked monthly by hand or with a tool	Anything statistically representative; treat it as a trend line

Check that the AI Assistant channel picks up every assistant you care about. If one is missing, create a custom channel group with an AI channel above Referral that matches its referrer.

Morrison pulls six months of AI assistant sessions from GA4, finds the landing pages they reach, checks which top queries show an AI Overview and asks Perplexity whether it cites the site.

A 30-day plan

Week 1, access. Decide your policy for search and training crawlers, update robots.txt, check CDN and firewall settings, confirm the crawlers in your logs, and check the Search generative AI setting.
Week 2, measurement. Check the AI Assistant channel in GA4, verify your site in Bing Webmaster Tools, look at Search Console’s generative AI report, and write your fixed list of questions.
Week 3, pages. Take the ten pages that answer your most valuable questions and check each against the “do” table: a plain answer early, original specifics, sources, dates.
Week 4, baseline. Record AI referrals, AI impressions and your prompt sample, so next quarter has something to compare with.

Where Morrison fits

Morrison connects to GA4, so chat can show sessions from AI assistants by month and by landing page, and to SerpApi for live Google results, including whether an AI Overview appears for your queries. Two Skills in the catalog are built for this work: the AI Search Traffic & Mention Tracker and the AI Crawler & Robots.txt Auditor. The page improvements themselves are ordinary content work: Morrison’s content analyzer scores every page in your site index, and chat writes the briefs page by page. See Skills and AI search visibility.

Key takeaways

Access first: each product cites from its own index. Allow its search crawler, and decide separately about training crawlers.
Google needs nothing special: indexed, snippet-eligible and the AI setting left on.
Be the source a summary needs: plain answers, original specifics, sources and honest dates.
Skip the shortcuts: llms.txt, chunking and fan-out pages do not help with Google, and some can hurt.
Measure with the new reports: Search Console’s AI report, Bing’s AI Performance and GA4’s AI Assistant channel.

Sources and further reading