GEO Research10 min read

ChatGPT and Perplexity Cite Different Sites: What Is Known About Source Overlap

Q: How do I know which platforms are citing me?

The direct way is to ask each platform questions about your topics and write down which sources it names. Causabi monitoring does this on paid plans: it asks ChatGPT, Gemini, Grok and Claude questions through their official APIs and reports whether your business is mentioned. Referral traffic in analytics shows only click-throughs, not every mention.

May 16, 2026Updated 11 October 2026

Free check: can AI read your site?

See your AI score before you read on

About 90 seconds · No email required · Free

GEO advice often treats AI platforms as interchangeable. According to a Profound study (published July 1, 2025; 100,000 prompts run on both ChatGPT and Perplexity), only 11.0% of domains are cited by both platforms. If so, the platforms mostly draw on different sets of sources. This is one vendor’s data from 2025: the figures may be different today.

What the data says

Profound, an AI-search monitoring vendor, ran 100,000 prompts on both ChatGPT and Perplexity and compared the domains in the answers. This is one vendor’s data from 2025, and the publication does not spell out how the share was calculated. A separate, wider Profound dataset — 680 million citations across ChatGPT, Google AI Overviews and Perplexity from August 2024 to June 2025 — is not the basis of this comparison.

Domains in ChatGPT and Perplexity answers (Profound, 100,000 prompts, July 1, 2025)

Domain cited by	Share
ChatGPT only	37.4%
Perplexity only	51.6%
Both	11.0%

If the 11% figure is right, the platforms probably pick sources in different ways. They do not publish how, and the descriptions below are partly assumptions.

ChatGPT: model knowledge and web search

ChatGPT has two sources of information. The first is what the model learned in training. The second is web search, which kicks in when current information is needed. What exactly the search uses (Bing’s index in particular) OpenAI does not describe in detail; this section relies on third-party descriptions and assumptions.

That suggests an assumption: sites that appear often in training data (Wikipedia, major publications, reference sites) may be mentioned more often, and this cannot be changed quickly. We have no confirmation from OpenAI.

What ChatGPT is said to favor

→ Wikipedia — per Profound (680M citations, August 2024 to June 2025), ChatGPT’s most-cited source: 7.8% of all citations and nearly half (47.9%) of citations among its ten most-cited sources. Wikipedia pages follow its own rules and exist only for notable organizations, so this is not something you can plan
→ Major editorial publications — Forbes, TechCrunch, Wired, trade press. Whether such mentions affect the model’s answers is unknown
→ Authoritative brand signals — Crunchbase and G2 profiles and official directory entries: they give a machine checkable facts about the company
→ Bing-indexed content — check that the site is not blocked for Bingbot; useful for ordinary search too

The ChatGPT strategy

Because a model’s knowledge updates slowly, the actions split into long-term and short-term.

Long-term: be mentioned where others write about you independently, such as trade publications and directories. How soon this shows up in model answers is unknown.

Short-term: make sure Bingbot and OAI-SearchBot are not blocked, submit your sitemap to Bing Webmaster Tools, and write copy that answers questions with current data.

Perplexity: real-time search and many links

Perplexity builds its answers on a web search and shows its sources next to the answer. How many links Perplexity gives per answer depends on the study: published estimates differ widely, so we do not quote a figure.

Some of the links may go to smaller, specialised sites. Whether such links bring visitors, we do not know.

What Perplexity is said to favor

→ Structured Q&A content — question-and-answer pages and FAQPage markup. Perplexity has not confirmed this
→ Fresh content — search runs in real time, so fresh pages may be found more easily. The “30 days” threshold is not confirmed anywhere
→ Blog posts and guides — articles written as an answer to a question
→ Perplexity bot access — check that PerplexityBot is not blocked in robots.txt
→ Specific numerical claims — pages with concrete figures and sources are easier to quote

The Perplexity strategy

Practical steps that do not depend on how exactly Perplexity chooses sources:

Keep material up to date and show real dates
Add FAQPage JSON-LD to pages that carry questions and answers
Structure articles to directly answer specific questions, not to cover topics broadly
Check robots.txt: PerplexityBot may be shut out by a rule written for all crawlers
Write for specific conversational questions, not only short keywords

Google AI Mode: many links to YouTube

According to Ahrefs (updated September 2, 2026, US queries across all topics), YouTube is the second most-cited domain in Google AI Mode with a 17.8% mention share, just behind Reddit (17.9%). Google has not given reasons; it may be linked to YouTube being owned by Google. That is our assumption.

Google Business Profiles and Knowledge Graph entries may also play a role, since Google has its own data on businesses. That is plausible, but Google has not confirmed it.

What Google AI Mode is said to favor

→ YouTube content (17.8% mention share per Ahrefs) — videos on YouTube with clear speech. Whether your business needs them is your call; we cannot promise a result
→ Google Business Profile — a verified, fully filled-in business profile is useful regardless of AI
→ Knowledge Panel presence — some brands have a Knowledge Panel. A link to AI Mode answers is not proven
→ Brand-owned site content — for facts about a company, its own site is the logical primary source
→ Schema.org markup — Organization, Product and FAQPage markup. Google has not said whether AI Mode uses it when choosing sources

The Google AI Mode strategy

If YouTube is that prominent in AI Mode answers, video can be part of GEO work. Even a simple screen recording with narration gives text Google can read. Whether it ends up in AI Mode answers cannot be said in advance.

Beyond video: fill in your Google Business Profile and Organization markup with current facts. This is useful whatever way AI Mode picks its sources.

See whether each engine mentions you

On paid plans Causabi asks ChatGPT, Gemini, Grok and Claude questions through their official APIs and reports whether your business is mentioned in each one’s answers. The free check takes about 90 seconds and shows your site’s AI score.

Check your AI score →

What you can do for each platform

What helps on every platform

Start here: this is the basis any AI service needs to read a site. No result is guaranteed.

AI bot access in robots.txt — verify GPTBot, ClaudeBot, PerplexityBot, Bingbot are not blocked

FAQPage JSON-LD on all pages with Q&A content

datePublished and dateModified in Article JSON-LD

Organization schema with complete NAP (name, address, phone) data

Content structured with direct, answer-first sections

Specific facts with a source

llms.txt at the root of the domain (a proposed format, not a standard)

ChatGPT-specific actions

If the organization meets Wikipedia’s notability criteria, independent editors may write the page; do not create it yourself

Tell trade publications about the product when you have news

Create or verify Crunchbase, G2, Capterra, and AlternativeTo profiles

Ensure Bingbot is not blocked and submit sitemap to Bing Webmaster Tools

Build citation-worthy content for current-information queries (recent stats, comparisons)

Perplexity-specific actions

Verify PerplexityBot is explicitly allowed in robots.txt

Update material when it goes out of date

Structure all editorial content around specific questions, not broad topics

Add FAQPage schema to every guide and blog post

Target long-tail, conversational query variants — not just head keywords

Google AI Mode-specific actions

If it suits your business, record videos with clear speech

Complete and verify Google Business Profile with current information

Check whether the brand has a Knowledge Panel and keep the Organization markup accurate

Complete Organization, Product, and SoftwareApplication schema on your site

Publish copy that answers customers’ questions about your category

What the 11% overlap means for your work

Most companies cannot run separate work for each platform. It is reasonable to start with the common basics in the list above, then look at your own data for where people come from and which AI services mention you. We have no ready-made proportions.

Frequently Asked Questions

Can you optimize for ChatGPT and Perplexity at the same time?

Yes. The basics are the same: AI bot access in robots.txt, clear copy, markup, current dates. The platforms do not publish how they choose sources, so differences between them remain assumptions.

Which AI platform should I focus on first?

Start from where your audience is. The platforms do not publish how they choose sources, so you can only compare their effect from your own data: monitoring and referral traffic show where people come from.

What about Claude citations?

Anthropic does not publish details of how Claude chooses sources when it uses web tools. The basics are the same: open access for bots, clear and precise copy, documentation.

Why does Google AI Mode cite YouTube so heavily?

According to Ahrefs (updated September 2, 2026), YouTube is the second most-cited domain in Google AI Mode with a 17.8% mention share (Reddit: 17.9%). Google has not said publicly why.

How do I know which platforms are citing me?

Ask each platform questions about your topics and write down which sources it names. Causabi monitoring does this on paid plans: it asks ChatGPT, Gemini, Grok and Claude questions through their official APIs and reports whether your business is mentioned. Referral data in analytics shows only click-throughs, not every mention.