Three engines, three different games
The most common mistake in AI visibility work is treating ChatGPT, Perplexity and Gemini as one channel. They are not. Conductor analyzed citations across seven answer engines over seven months, from September 2025 to March 2026, and found that only 11% of cited domains appeared on more than one platform. Each engine crawls differently, indexes differently and trusts different corners of the web. A playbook that works must start from how each system actually sources its answers, so that is where this one starts.
How ChatGPT picks sources
OpenAI operates three crawlers with separate robots.txt tokens, and confusing them costs visibility. GPTBot gathers training data. OAI-SearchBot builds the index behind ChatGPT search and its citations. ChatGPT-User fetches pages live when a user asks about them. Blocking GPTBot to keep your content out of training does not remove you from ChatGPT search, but blocking OAI-SearchBot does.
ChatGPT search has historically leaned on Bing: Seer Interactive found in early 2025 that 87% of its citations matched Bing top results, so Bing Webmaster Tools deserves the attention most teams only give Google. Profound's analysis of 680 million citations found ChatGPT's most cited source by far is Wikipedia, followed by Reddit and established media, a picture consistent with the licensing deals OpenAI signed with publishers like News Corp, Axel Springer, the Financial Times and AP.
How Perplexity and Gemini pick sources
Perplexity runs retrieval on its own index, built by PerplexityBot and deliberately curated toward high quality sources rather than web scale. Its citation diet is distinct: in Profound's data, Reddit is Perplexity's most cited domain, with YouTube second. Community proof matters more there than anywhere else. Gemini and Google AI Overviews are grounded in Google Search itself, which reduces most of the Gemini problem to familiar territory: if you are indexed and competitive in Google, you are eligible.
One technicality worth knowing: Google-Extended is a robots.txt token that controls AI training and Gemini grounding, not a crawler, and blocking it does not affect your Search rankings or AI Overview eligibility.
11%
of cited domains appear on more than one AI platform (Conductor, 7 month study)
0
major AI crawlers execute JavaScript when crawling (Vercel and Merj, December 2024)
~7 days
median time to a first ChatGPT citation for new content (Profound benchmark, 2026)
+41%
visibility from adding quotations, the strongest content tactic measured (KDD 2024)
The technical gate almost everyone fails
Vercel and Merj studied AI crawler behavior in December 2024 and found that none of the major AI crawlers execute JavaScript. GPTBot downloads your JS files but never runs them. If your content only exists after client side rendering, ChatGPT and Perplexity see an empty page. Server side rendering is the single highest leverage technical fix for AI visibility. The exception is Gemini, which inherits Googlebot's rendering.
Content that earns citations, per the data
The Princeton GEO study (KDD 2024) measured nine tactics across 10,000 queries: adding quotations lifted visibility about 41%, statistics about 32% and cited sources about 30%, results the authors partially validated on real Perplexity queries, where adding quotations improved position adjusted visibility by 22%. Format matters too. HubSpot's State of AEO research and Wix's AI Search Lab, covering more than a million citations combined, found listicles earn 21.9% of citations, articles 16.7% and product pages 13.7%, with comparison content performing best on ChatGPT specifically.
The practical translation: structured, opinionated comparisons and evidence dense articles are citation magnets; thin brochure pages are invisible. And place the answer early: engines quote passages, and analysis of ChatGPT citations suggests a large share pull from the first third of the content.
The playbook, in order of leverage
- Verify robots.txt allows OAI-SearchBot, ChatGPT-User, PerplexityBot and GPTBot, each a deliberate decision
- Confirm your content renders without JavaScript and your site is indexed in Bing Webmaster Tools
- Rewrite your ten most valuable pages so each section opens with a direct, quotable answer
- Add named statistics, expert quotes and source citations to every page that matters
- Publish comparison content in your category, the format AI engines cite most for commercial intent
- Build presence on Reddit, YouTube and review platforms, the sources Perplexity and AI Overviews trust most
- Earn coverage in established media and Wikipedia where legitimate, the sources ChatGPT trusts most
How fast it works and how to measure it
This channel moves faster than SEO ever did. Profound's benchmark from January 2026 found the median time to a first ChatGPT citation for new content was under seven days, with 90% of first citations arriving within 37 days. Semrush found Google AI Mode can cite content within 24 hours. But citations are volatile: Yotpo measured only a 15% chance that a URL cited by ChatGPT today still appears in the same answer three weeks later, because answers are generated fresh each time. So measure share of voice, not individual placements.
The lightweight version: a fixed set of 10 to 30 real buyer questions, asked monthly in each engine, logged in a spreadsheet. The tooled version: platforms like Profound, Otterly, Peec AI, Semrush AI Visibility or Ahrefs Brand Radar, plus watching GA4 referrals from chatgpt.com and perplexity.ai and AI crawler hits in your server logs.
What nobody can promise you
Honesty clause: no vendor knows the ranking criteria of these engines, because none are published. Every citation study is an observational snapshot of a fast moving target, and the correlation between brand mentions and visibility does not prove causation. llms.txt, the proposed standard for AI crawlers, had near zero real usage in log analyses covering hundreds of millions of AI bot visits, so add it if you like, but expect nothing from it.
What the evidence does support is direction: retrievable, evidence dense, answer shaped content, plus presence on the platforms each engine already trusts, measurably increases your odds of being the recommendation. In a channel this young, being early and consistent is the closest thing to an unfair advantage.
Only 11% of domains cited by one AI engine are cited by another. There is no generic AI visibility, only engine by engine presence.
Written by
Alexander Makeev
CTO
Full-stack engineer and AI specialist with a background in nuclear physics and experience at CERN. Expert in Python, cloud infrastructure and LLM integration. He architects the technical systems behind our clients' automation and AI-search work.
