12 ways to get your site cited in AI answers (2027)
Allow each assistant's search crawlers, and make sure your firewall serves them the real page with the text in the HTML. Then answer the buyer's question in the first lines, state facts in plain text, add first-hand detail, show who publishes and when, use matching JSON-LD, and be listed truthfully on other sites and in product feeds.
No vendor sells a citation and no one can promise one. What the vendors do publish is how their crawlers reach pages, what they need to read them, and what they look at when they pick sources. This list collects those documented points in the order to work through them. The first four decide whether an assistant can use your page at all; the rest decide whether your page is the best answer it finds.
On this page
- Which of these ways matter most, at a glance?
- 1. Allow the search and user crawlers in robots.txt
- 2. Make your firewall serve AI crawlers the same page as a browser
- 3. Put the main text in the HTML your server sends
- 4. Keep pages indexed and eligible for a snippet in Google
- 5. Answer the question in the first lines of the page
- 6. Use the questions your buyers ask as headings
- 7. Publish something original that others cannot copy
- 8. State prices, specs and locations in plain text
- 9. Show who publishes the page and when it changed
- 10. Add JSON-LD that matches the visible text
- 11. Be described accurately on other sites
- 12. Keep product feeds and business profiles current
- How we chose these, and what we left out
Which of these ways matter most, at a glance?
Work from the top. A later item cannot help while an earlier one fails, and the first three take minutes to check.
- 1. Allow the search crawlers: effort low, about 15 minutes, needed for every assistant.
- 2. Let your firewall serve them the page: effort low to medium, needed if you use a CDN or bot protection.
- 3. Text in the initial HTML: effort low to check, high to fix on a client-side app.
- 4. Indexed and snippet-eligible: effort low, needed for Google AI Overviews and AI Mode.
- 5. Answer in the first lines: effort low per page, applies to every page you want cited.
- 6. Buyers' questions as headings: effort low per page.
- 7. Something original: effort high, the main reason to cite you over others.
- 8. Facts in plain text: effort low, matters most for prices, specs and locations.
- 9. Publisher and dates: effort low, one template change.
- 10. Matching JSON-LD: effort medium, one template per page type.
- 11. Honest outside listings: effort high and slow, outside your full control.
- 12. Product and local data feeds: effort medium, only for shops and local businesses.
1. Allow the search and user crawlers in robots.txt
Each assistant reads the web through its own agents, and robots.txt decides which of them may crawl. If the search crawler is disallowed, the assistant has nothing of yours to cite.
The tokens to allow are OAI-SearchBot for ChatGPT search, Claude-SearchBot and Claude-User for Claude, PerplexityBot for Perplexity and Googlebot for Google's AI Overviews and AI Mode. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, though they can still appear as navigational links. Training crawlers such as GPTBot and ClaudeBot are separate tokens, so you can block training and still allow search. In our study of the 1,000 most popular domains, 19.5% of the 518 sites that served a robots.txt blocked at least one AI search or user crawler from the home page.
A group that names a crawler replaces the User-agent: * group for that crawler, so check every named group, not only the wildcard. OpenAI and Perplexity both say changes can take about 24 hours to reach their systems.
Best for: every site. Our guide to robots.txt for AI crawlers has full examples, and the guide on blocking AI training while staying in AI search shows the split.
2. Make your firewall serve AI crawlers the same page as a browser
robots.txt is only a request. A CDN, WAF or bot protection can still answer a crawler with a 403, a 429 or a challenge page while your own browser sees the page normally.
Anthropic says its bots will not try to bypass CAPTCHAs, so a challenge ends the visit. OpenAI, Anthropic and Perplexity publish the IP addresses of their agents, at openai.com/searchbot.json, claude.com/crawling/bots.json and perplexity.ai/perplexitybot.json. Allow those ranges rather than trusting the user agent string alone, which anyone can fake.
Skip if: your site has no CDN, firewall or bot protection at all. Even then, check once after your next hosting change. Our guide on Cloudflare blocking AI crawlers covers the most common case.
3. Put the main text in the HTML your server sends
A crawler that does not run JavaScript sees only the HTML the server returns. If your answer is added later by scripts, that crawler gets an empty page.
Google renders JavaScript for Search, but still recommends server-side rendering or prerendering because not all bots can run it. OpenAI, Anthropic and Perplexity do not say in their crawler documentation that their agents execute JavaScript, so the safe assumption is that they read the initial HTML only. Test it by downloading the raw HTML and searching it for a sentence from your answer: curl -s https://example.com/page/ | grep -c "a sentence from the page". A result of 0 means the text is not there.
Skip if: your pages come from a classic CMS or a static site generator; they usually pass. Single-page apps should read our guide on JavaScript content and AI crawlers.
4. Keep pages indexed and eligible for a snippet in Google
Google's AI Overviews and AI Mode are built on Google Search. Google says a page must be indexed and eligible to be shown with a snippet, and that there are no additional technical requirements.
So the usual blockers apply: a noindex left over from staging, a nosnippet rule, or a very small max-snippet value. Google's robots meta tag documentation says nosnippet also stops the content from being used as a direct input for AI Overviews and AI Mode. Use URL Inspection in Search Console to confirm the page is indexed. For Microsoft Copilot, Bing Webmaster Tools has an AI Performance report that shows how often your pages are cited.
Best for: any site that wants to appear in Google's AI features. Our article on Google-Extended explains why blocking that token does not remove you from AI Overviews.
5. Answer the question in the first lines of the page
Put the direct answer in two or three sentences right under the H1, then explain. An assistant that compares several sources finds your answer without reading the whole page, and a reader gets it at once.
Google's guide to generative AI features says you do not need to break content into tiny pieces or rewrite it for AI, because its systems understand synonyms and long pages with several topics. Answer-first is not a trick for machines; it is a clear page for people that also happens to be easy to quote. Write the answer as a complete statement that still makes sense when quoted alone, with the subject named instead of "it" or "this".
Best for: guides, product pages, pricing pages and FAQs. Skip if: the page is a tool or a form, where the tool itself is the answer.
6. Use the questions your buyers ask as headings
Headings that match real questions tell a reader and a crawler what each part of the page answers. "How much does it cost per user?" is easier to match to a question than "Pricing philosophy".
List the 10 to 30 questions buyers ask before they know your name, the way they phrase them, and give each one a heading with a direct answer under it. Do not stuff extra keyword variants into the text: Google says its systems understand meaning and you do not need extra long-tail keywords for AI search.
Best for: long guides and comparison pages. Skip if: a short page answers one question; the H1 already does this job.
7. Publish something original that others cannot copy
An assistant has little reason to cite a page that repeats what ten other pages say. Google's guide warns against commodity content built on common knowledge and asks for a unique point of view, such as first-hand experience.
Original material includes your own measurements, test results, prices you collected, real customer questions with your answers, screenshots of the steps, and data from your product. Google's helpful content guidance asks whether a page provides original information, reporting, research or analysis. Say how you got the numbers and when, and link to the primary sources behind any claim you did not measure yourself.
Best for: sites that compete with many similar pages. Skip if: you have nothing original yet; improve items 1 to 6 first and collect data as you go.
8. State prices, specs and locations in plain text
Facts locked in images, PDFs, sliders or tabs that load later are hard for any crawler to read. Write them as text on the page, with units and currency.
"From $19 per user a month, billed yearly" can be quoted. A price that appears only inside a calculator widget cannot. The same goes for opening hours, supported countries, file size limits and version numbers. Google lists keeping important content in text form among its best practices for AI features.
Best for: pricing, product, service and location pages.
9. Show who publishes the page and when it changed
Readers and assistants both weigh who stands behind a claim and how current it is. Make both visible.
Show a labeled date such as "Updated 19 November 2026" and put the same dates in datePublished and dateModified in your Article or BlogPosting markup; Google asks for the visible and marked-up dates to agree. Name the publisher and link to an about page. Do not change dates without changing the content: Google's helpful content guidance lists changing dates to make pages seem fresh as a warning sign.
Best for: articles, guides and documentation.
10. Add JSON-LD that matches the visible text
Structured data states in machine-readable form what the page is about: the organization, the article, the product and its price. It removes guesswork, but it is not a switch.
Google says no special schema is needed for its AI features and that structured data should match the visible text. Use Organization once, with sameAs links to your official profiles, and Article, Product or SoftwareApplication on each page. Markup that claims a price or a rating the page does not show is a problem, not a boost. Test the result in the Rich Results Test or the Schema Markup Validator.
Best for: sites with products, prices or many articles. Our guide to structured data for AI answers has copy-ready examples.
11. Be described accurately on other sites
Assistants do not only read your own pages. When they recommend a product or a business, they often cite reviews, directories, comparison articles and forum threads, so your absence there is a gap no change on your site can fill.
OpenAI says that when ChatGPT chooses products it considers structured metadata from first-party and third-party providers and other third-party content. Make sure the listings that exist describe you correctly and consistently: name, category, price range, location. Do not buy or fake mentions. Google's guide says that seeking inauthentic mentions across the web is not as helpful as it may seem, and fake reviews break most platforms' rules.
Best for: brands, products and local businesses. Our article on why AI recommends a competitor shows how to find the sources assistants cite in your category.
12. Keep product feeds and business profiles current
Some answers are built from structured catalogs rather than from crawled pages. For shops and local businesses those catalogs are a direct way in.
Google says Merchant Center feeds can make products visible in AI responses and that Business Profile information can do the same for local businesses and services. OpenAI lets businesses submit product feeds for ChatGPT product discovery through its Business Hub. Keep prices and stock in these feeds the same as on your site, because a mismatch gives the assistant two answers to choose from.
Skip if: you sell no physical products and have no location customers visit. Our list of ways to get your products into AI answers covers this in detail.
How we chose these, and what we left out
Every item is based on what OpenAI, Anthropic, Perplexity and Google document about their crawlers and answer features, or on Google's published guidance for Search. Where a vendor documents nothing, for example how Claude or Perplexity rank sources, we say so instead of guessing. The figures on blocked crawlers come from our own study of the 1,000 most popular domains.
We left out tactics the vendors call unhelpful or that have no documented effect: chopping pages into small chunks, writing separate versions for AI, buying mentions, and counting on llms.txt for Google, which Google says its Search does not use. Publishing llms.txt does no harm and may help other tools; see our guide on llms.txt. None of the items guarantees a citation. They make one possible and give the assistant a reason to pick your page.
AISeeWhy checks items 1 to 4 for you: it fetches a page as 16 AI and search crawlers next to a real browser, applies robots.txt per crawler, spots challenge pages and shows how much text arrives without JavaScript.
Fixes for this problem
- Long paragraphs: 2
- Sections do not open with a short answer: 2 of 3
- No clear next step on a commercial page
- No links to outside sources
Next steps
- robots.txt tester and checker for AI crawlers
- llms.txt checker and validator
- See an example full report
- AI crawler monitoring
- Plans and prices
Crawlers in this article
Frequently asked questions
How long does it take to get cited after fixing access?
OpenAI and Perplexity say robots.txt changes can take about 24 hours to reach their systems. Google says changes to snippet controls can take days to months until pages are recrawled. After that, a citation depends on whether your page is the best source for a question, so there is no fixed delay.
Can I pay to be cited in ChatGPT or Perplexity answers?
No vendor documents a way to buy a citation in an answer. Ads and product feeds are separate programs with their own rules. The organic route is the one in this list: let the crawlers in and publish the clearest, most original answer.
Do I need llms.txt to be cited?
No. Google says its Search does not use llms.txt, and OpenAI, Anthropic and Perplexity do not list it as a requirement for their crawlers. It is optional and harmless. Crawler access, text in the HTML and clear answers matter far more.
Does blocking GPTBot or ClaudeBot stop my site from being cited?
Not by itself. GPTBot and ClaudeBot collect content for model training. Citations in ChatGPT search depend on OAI-SearchBot, and in Claude on Claude-SearchBot and Claude-User. Keep training and search crawlers in separate robots.txt groups so a training block does not catch the search crawlers.
Sources
- OpenAI: Overview of OpenAI crawlers
- Anthropic: Does Anthropic crawl data from the web, and how can site owners block the crawler?
- Perplexity: Perplexity crawlers
- Google Search Central: AI features and your website
- Google Search Central: Optimizing your website for generative AI features on Google Search
- Google Search Central: Robots meta tag, data-nosnippet and X-Robots-Tag
- Google Search Central: Creating helpful, reliable, people-first content
- Google Search Central: Influence your byline dates
- Google Search Central: JavaScript SEO basics
- Google Search Central: Organization structured data
- OpenAI: Shopping with ChatGPT Search
- OpenAI: Submit product feeds
- Microsoft Bing: Introducing AI Performance in Bing Webmaster Tools Public Preview
- AISeeWhy: AI Crawler Access Report, 2026-11