Sketch-style illustration of a staircase labelled Authority, Relevance, Clarity, Original Insight, Technical Quality and Trust & Signals leading up to the ChatGPT and Perplexity logos

How to Get Cited by ChatGPT and Perplexity: What Each Engine Actually Uses

To get cited by ChatGPT and Perplexity, start by making sure their search crawlers can reach your pages: OAI-SearchBot for ChatGPT and PerplexityBot for Perplexity. Then publish pages where each section answers one question with specific, sourced facts, cover the related questions people ask around your topic, and earn mentions of your brand on other sites. Finally, measure it, because neither engine tells you when it cites you. None of this guarantees a citation. It removes the reasons you get skipped.

ChatGPT and Perplexity don't find sources the same way

Treating "AI search" as one channel is a common mistake. The two engines retrieve, select and cite pages through different systems, so the same page can be cited by one and ignored by the other.

ChatGPT doesn't search the web for every prompt. Ahrefs reports that its citation behaviour is query-dependent, with internal systems deciding when a search is needed [4]. When it does search, OpenAI's OAI-SearchBot is the crawler that surfaces websites in ChatGPT's search features, and OpenAI states that sites opted out of it will not be shown in ChatGPT search answers [1]. OpenAI also notes it can obtain URLs from a third-party search provider [2]; Ahrefs describes Bing as that built-in provider, with independent tests suggesting Google is used as a fallback [4].

Perplexity was built to cite and searches the web in real time for its answers. It runs its own search index, built by its crawler PerplexityBot, rather than drawing on Google's or Bing's [3][4].

The difference shows up in the data. In an Ahrefs study of 15,000 long-tail prompts (data collected in July 2025), 28.6% of the URLs Perplexity cited ranked in Google's top 10 for the same query. For ChatGPT and the other assistants tested, the figure hovered around 8%, and more than 80% of their citations came from pages that did not rank for the query at all [4]. In practice, Perplexity tends to reward pages that already rank well, while ChatGPT draws from a wider and less predictable pool.

FeatureChatGPTPerplexity
When it searchesOnly when its systems decide a prompt needs itSearches for its answers in real time
Where results come fromOAI-SearchBot plus a third-party search providerIts own index, built by PerplexityBot
Crawler to allowOAI-SearchBot (GPTBot is for model training)PerplexityBot
Cited URLs in Google's top 10 (Ahrefs, July 2025 data)About 8%28.6%
How clicks show up in analyticsutm_source=chatgpt.com on referral linksperplexity.ai referrals

Step 1: Let the right crawlers in

OpenAI runs separate user agents for separate jobs: GPTBot for model training, OAI-SearchBot for search, and ChatGPT-User for certain actions users trigger inside ChatGPT. The settings are independent, so you can block training while staying visible in search [1]. Perplexity's indexing crawler is PerplexityBot, and you control it through robots.txt like any other bot [3].

A robots.txt that keeps you citable in both engines while opting out of OpenAI model training looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: GPTBot
Disallow: /

Drop the GPTBot block if you're happy for your content to be used in training. Either way, blocking GPTBot does not remove you from ChatGPT search.

Two things robots.txt won't reveal. First, a CDN or firewall rule can block these bots even when robots.txt allows them, so check your server logs for the user agents. OpenAI publishes IP ranges for each bot, which lets you confirm a visit is genuine [1]. Second, rendering: a December 2024 crawl-log study by Vercel and MERJ found that none of the major AI crawlers it measured executed JavaScript, even though OpenAI's crawler downloaded JavaScript files [5]. If your main content only appears after client-side JavaScript runs, these crawlers may see an empty page. Our comparison of static generation and server-side rendering in Next.js covers the rendering options.

Step 2: Write sections an engine can lift and trust

The strongest controlled evidence on what makes content more visible in AI answers comes from the GEO study by researchers at Princeton and IIT Delhi, presented at KDD 2024. Adding citations to credible sources, quotations and statistics produced relative improvements of 30–40% on the study's main visibility metric. Making text more fluent and easier to understand added 15–30%. A more persuasive, authoritative tone produced no significant gain, and keyword stuffing offered little to no improvement; tested on Perplexity, it performed 10% worse than doing nothing [6].

One finding matters for smaller sites. The gains were largest for pages that ranked lower in search: citing sources raised the visibility of the fifth-ranked source by 115.1%, while the top-ranked source lost 30.3% on average [6]. Treat the direction as more reliable than the size, though. The experiments used GPT-3.5 and the top five Google results in 2023–24, and today's engines work differently.

This doesn't contradict Google's guidance that you don't need to rewrite content "just for AI" or break it into tiny chunks [7]. The gains in the study came from adding evidence a reader can check, not from formatting tricks. In practice:

  • Answer the heading's question in the first sentence or two of each section, then add the nuance.
  • Make each section stand on its own. Name the subject instead of writing "it" or "this tool", so a passage still makes sense when it's quoted without the rest of the page.
  • Replace vague claims with numbers, and say where each number comes from and when it was measured.
  • Quote named, credible sources where their exact words carry weight.
  • Add something only you can publish: client results, test data, a decision framework built from your own work. Google calls this non-commodity content, and it's the thing an engine can't get from the ten other pages answering the same question [7]. Our guide to writing a credible case study shows how to present that kind of evidence.

Step 3: Cover the questions around your topic, not one keyword

AI engines rarely search for exactly what the user typed. Google describes query fan-out as a set of related queries the model generates to fetch more results; a question about weedy lawns might fan out into searches about herbicides, chemical-free removal and prevention [7]. Ahrefs argues that AI assistants retrieve pages the same way and favour pages that show up consistently across the variations, so a page ranking around sixth for three related queries can be cited over a page ranking first for only one [4].

That's also why "which prompts lead to citations?" has a less satisfying answer than people hope. The prompt is a starting point, and the engine searches its own rewrites of it. For ChatGPT, Ahrefs' reading is that a search is triggered when a prompt is hard to answer from training data alone [4], which means prompts about recent, specific or niche topics are where a citation is most likely to be on offer.

We can see traces of this in the Search Console data for this very article. Between July and early October 2026, it collected impressions for queries that read like full AI prompts, including one that opens with a location instruction followed by "question:". We can't confirm what generated them, but they look more like machine-written queries than anything a person types into Google.

The practical move is to list the questions that surround your topic, by asking ChatGPT and Perplexity yourself and noting the follow-ups they suggest, and then answer them in clear sections on one strong page. Don't spin up a separate thin page for every variation. Google warns that creating content for every fan-out query mainly to manipulate rankings or AI responses violates its scaled content abuse policy [7].

Step 4: Earn mentions beyond your own site

What other sites say about you appears to matter as much as what you publish. In an Ahrefs study of 75,000 brands, YouTube mentions had the strongest correlation with brand visibility across ChatGPT, Google AI Mode and AI Overviews (about 0.737), and branded web mentions followed closely at 0.66–0.71 [8]. Two caveats: this is correlation rather than proof of cause, and Perplexity wasn't one of the platforms studied.

Google is blunt that chasing inauthentic mentions doesn't help, because its spam systems sit underneath its AI features [7]. The mentions worth earning are the ones a buyer would trust anyway: coverage in industry publications, appearances on podcasts and YouTube channels your audience watches, listings on the review and comparison sites in your category, and genuinely useful answers in the communities where your customers ask questions.

Step 5: Track whether you're actually being cited

Neither ChatGPT nor Perplexity offers a dashboard of your citations, so measurement means combining a few imperfect sources.

MethodChatGPTPerplexityWhat it can't tell you
Manual prompt panelYesYesAnything at scale; results vary between runs
Referral analytics (GA4 or similar)YesYesCitations that didn't earn a click
Bing Webmaster Tools AI PerformanceNoNoCovers Copilot and Bing only
Ahrefs Brand RadarYesYesYour exact prompts unless you add custom ones
Semrush AI Visibility ToolkitYesYesEvery phrasing your buyers actually use

A manual prompt panel costs nothing. Write 20–30 prompts your buyers would realistically ask, run them in ChatGPT (with search on) and Perplexity every month, and record whether you're cited and who is cited instead. Run each prompt more than once: Ahrefs notes that the same query can produce different citations depending on the user's prompt wording and conversation history [4].

Referral analytics show which citations earn clicks. OpenAI says ChatGPT automatically adds utm_source=chatgpt.com to referral URLs [2], so you can filter for it in Google Analytics. For Perplexity, look for sessions referred from perplexity.ai.

Bing Webmaster Tools' AI Performance report counts citations in Copilot and Bing's AI summaries and shows the fan-out queries behind them, but it does not cover ChatGPT [9]. It's still the closest thing to first-party citation data for a chat assistant. Google Search Console's Generative AI performance report does the equivalent job for AI Overviews and AI Mode [7].

Paid trackers sample AI answers at scale. Ahrefs Brand Radar runs search-backed prompts through ChatGPT, Perplexity, Gemini, Copilot, AI Overviews and AI Mode, and lets you add custom prompts [10]. Semrush's AI Visibility Toolkit tracks prompts across ChatGPT, Perplexity, Gemini, AI Overviews, AI Mode and other assistants [11]. Treat their numbers as samples, not counts: AI answers change from session to session, and no tool sees every conversation.

What you can skip

  • llms.txt files. Google says Google Search ignores them [7], and we found no OpenAI or Perplexity documentation saying their crawlers use them either.
  • Keyword stuffing. It hurt visibility on Perplexity in the GEO study [6].
  • Special schema for AI. Google says no special markup is required for its generative AI features, though structured data still helps with rich results [7].
  • A page for every prompt variation. One thorough page beats twenty thin ones, and Google treats the thin-page approach as spam [7].

Frequently asked questions

Do I need to rank on Google to get cited by ChatGPT or Perplexity?

It helps more with Perplexity than with ChatGPT. In Ahrefs' data, 28.6% of Perplexity's citations ranked in Google's top 10 for the query, against around 8% for ChatGPT [4]. Strong rankings make a Perplexity citation more likely; for ChatGPT, broad coverage of related questions and off-site mentions appear to matter more.

If I block GPTBot, will I disappear from ChatGPT?

No. GPTBot governs model training, while OAI-SearchBot governs whether you appear in ChatGPT search answers, and OpenAI treats the two settings independently [1].

Can a small or new site get cited?

Yes, and the evidence suggests smaller sites have the most to gain. In the GEO study, pages ranked lower in search saw the largest visibility gains from adding sources, quotations and statistics [6]. A small site still needs to be crawlable and to offer something the bigger pages don't.

How do I know whether ChatGPT is citing my content?

Run your key prompts in ChatGPT with search enabled, check your analytics for sessions tagged utm_source=chatgpt.com [2], and use a tracker such as Ahrefs Brand Radar or Semrush's AI Visibility Toolkit if you need visibility across many prompts.

Where to start

If you do one thing this week, check robots.txt and your firewall rules for OAI-SearchBot and PerplexityBot; nothing else on this list works if the crawlers can't reach you. Then pick your five most commercially important pages, rewrite their key sections so each one answers a question with sourced specifics, and start a monthly prompt panel so you can see whether anything changes.

If you'd rather have a team run the crawler audit, content rework and citation tracking for you, Website Vikreta's digital marketing team handles SEO and GEO for businesses that want to be found in both search results and AI answers.

References

  1. OpenAI — Overview of OpenAI Crawlers
  2. OpenAI Help Center — Publishers and Developers FAQ
  3. Perplexity — PerplexityBot
  4. Ahrefs — Only 12% of AI Cited URLs Rank in Google's Top 10 for the Original Prompt (published August 2025; data collected July 2025)
  5. Vercel — The Rise of the AI Crawler (December 2024)
  6. Aggarwal et al. — GEO: Generative Engine Optimization (KDD 2024)
  7. Google Search Central — Optimizing your website for generative AI features on Google Search
  8. Ahrefs — Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied)
  9. Adobe Experience League — Bing Webmaster Tools AI Performance Report
  10. Ahrefs — Brand Radar
  11. Semrush — Best AI Visibility Tools

Comments

Be the first to comment.

Related reads

Ready when you are.

Free call. No pitch. Tell us what's broken, and we'll tell you what we'd do.

Website Vikreta
© 2026 Website Vikreta. All rights reserved.Designed & Developed with AI-first precision