Parse
Work with usPricing
Sign inCheck your brand
Research/ChatGPT Search vs Google AI Mode: same brands, different sources

ChatGPT Search vs Google AI Mode: same brands, different sources

The two newest AI search surfaces do not read the same web. On the same 18,206 prompts, ChatGPT Search and Google AI Mode share only 6.2% of cited sources per prompt, while sharing 26.9% of named brands.

By Dimitry Apollonsky · July 8, 2026 · 8 min read

Overlap on matched prompts
  • Named brands26.9%
  • Cited sources6.2%
The engines are much closer on the brands they name than on the sources they cite.
▸Contents
  • The engines agree more on brands than on sources
  • Google AI Mode is the source-heavy surface
  • The top sources are different lists
  • YouTube is an AI Mode source, not a ChatGPT Search source
  • AI Mode leans into Google-owned and social surfaces
  • ChatGPT Search leans more toward vendor and specialist pages
  • The top brands mostly overlap anyway
  • The gap in sources cited held every week
  • The new surfaces split sources harder than the old pair
  • The operating rule: split source work by surface
  • Get the data
  • Sources
  • Related research
Contents
  • The engines agree more on brands than on sources
  • Google AI Mode is the source-heavy surface
  • The top sources are different lists
  • YouTube is an AI Mode source, not a ChatGPT Search source
  • AI Mode leans into Google-owned and social surfaces
  • ChatGPT Search leans more toward vendor and specialist pages
  • The top brands mostly overlap anyway
  • The gap in sources cited held every week
  • The new surfaces split sources harder than the old pair
  • The operating rule: split source work by surface
  • Get the data
  • Sources
  • Related research

We compared 18,206 matched prompts, meaning the same prompts run on both ChatGPT Search and Google AI Mode, from May 24 to June 25, 2026, covering 220,896 answers, 1.92M citations, and 46 industry groups.

6.2%
average cited-source overlap per prompt
26.9%
average named-brand overlap per prompt
14.0 vs 3.7
domains cited per answer
AI Mode vs ChatGPT Search
18,206
matched prompts

The engines agree more on brands than on sources

ChatGPT logoChatGPT Search and Google logoGoogle AI Mode do not read the same set of sources. On identical prompts, the average cited-source overlap is just 6.2%. The average named-brand overlap is 26.9%.

That means the engines often arrive at a similar shortlist through different evidence. Optimizing for AI search as one blended channel misses the real difference: brand presence and source presence are diverging.

6.2%
Cited-source overlap
26.9%
Named-brand overlap
2.89
Shared domains per prompt

Takeaway

The split is not only ChatGPT logoChatGPT versus Google logoGoogle. The two engines agree on the answer and disagree on the evidence.

Google AI Mode is the source-heavy surface

Google logoGoogle AI Mode cited at least one source in 99.3% of answers and averaged 14.0 distinct domains per answer. ChatGPT logoChatGPT Search cited in 65.6% of answers and averaged 3.7 domains. Per prompt, AI Mode exposed 42.4 distinct domains versus 12.1 on ChatGPT logoChatGPT Search.

  • AI Mode: domains per answer14.0
  • ChatGPT logoChatGPT Search: domains per answer3.7
  • AI Mode: domains per prompt42.4
  • ChatGPT logoChatGPT Search: domains per prompt12.1
How many domains each surface cites.

The top sources are different lists

ChatGPT logoChatGPT Search starts with Reddit logoReddit, then tech-review, academic, news, encyclopedia, and developer sources. Google logoGoogle AI Mode starts with YouTube logoYouTube, Google logoGoogle.com, Reddit logoReddit, LinkedIn logoLinkedIn, Facebook logoFacebook, Medium, and Instagram logoInstagram. Reddit logoReddit is the only source in both top three.

Top cited sources by surface
1reddit.com faviconreddit.com7.11youtube.com faviconyoutube.com5.05
2techradar.com favicontechradar.com1.34google.com favicongoogle.com4.92
3arxiv.org faviconarxiv.org1.03reddit.com faviconreddit.com4.15
4forbes.com faviconforbes.com0.6linkedin.com faviconlinkedin.com1.05
5en.wikipedia.org faviconen.wikipedia.org0.52facebook.com faviconfacebook.com1.02
6github.com favicongithub.com0.32medium.com faviconmedium.com0.95

YouTube is an AI Mode source, not a ChatGPT Search source

The largest individual split is YouTube logoYouTube. It supplies 5.05% of Google logoGoogle AI Mode citations and only 0.11% of ChatGPT logoChatGPT Search citations, about a 48x lean toward AI Mode. Facebook logoFacebook shows an even larger lean, while TechRadar and arXiv lean hard toward ChatGPT logoChatGPT Search.

Sources with the largest surface lean
facebook.com faviconfacebook.com0.0161.023Google logoGoogle AI Mode~64x1
techradar.com favicontechradar.com1.3420.024ChatGPT logoChatGPT Search~56x2
youtube.com faviconyoutube.com0.1065.054Google logoGoogle AI Mode~48x3
arxiv.org faviconarxiv.org1.0280.1ChatGPT logoChatGPT Search~10x4
linkedin.com faviconlinkedin.com0.1311.045Google logoGoogle AI Mode~8x5
en.wikipedia.org faviconen.wikipedia.org0.5190.07ChatGPT logoChatGPT Search~7x6

AI Mode leans into Google-owned and social surfaces

Google logoGoogle AI Mode's two most-cited sources are YouTube logoYouTube and Google logoGoogle.com, together about 10% of its citations. Its user-generated and social platform group reaches 12.3%, versus 7.4% on ChatGPT logoChatGPT Search.

  • Google logoGoogle AI Mode12.3%
  • ChatGPT logoChatGPT Search7.4%
Share of citations from major user-generated and social platforms.

ChatGPT Search leans more toward vendor and specialist pages

ChatGPT logoChatGPT Search sends 26.6% of citations to vendor or SaaS pages, compared with 17.3% for Google logoGoogle AI Mode. It also leans more toward editorial review pages and academic sources. That makes owned technical/product pages more important on ChatGPT logoChatGPT Search than the social-heavy AI Mode source mix implies.

What kinds of sources each surface cites
Vendor/site26.6217.3
General web22.2532.07
Social7.7312.39
Editorial6.343.6
Directory2.941.86
Academic1.170.29

The top brands mostly overlap anyway

Despite the evidence split, the top brand lists are similar. Datadog logoDatadog ranks first on both, and roughly 13 of each surface's top 15 brands overlap, including ClickUp logoClickUp, Jira, Asana logoAsana, Monday logoMonday, Grafana, Microsoft logoMicrosoft, Google logoGoogle, Amazon logoAmazon, Semrush logoSemrush, LinkedIn logoLinkedIn, Notion logoNotion, and Linear logoLinear.

13 of 15
top brand names overlap across the two surfaces

Takeaway

The engines often agree on who belongs in the answer while disagreeing on why.

The gap in sources cited held every week

The gap in how many sources each engine cited was not one unusual week. Across five full weeks, ChatGPT logoChatGPT Search stayed between 3.4 and 4.3 domains per answer, while Google logoGoogle AI Mode stayed between 12.1 and 18.7.

Domains cited per answer, by week
Jun 13.6414.67
Jun 83.3812.1
Jun 153.5413.47
Jun 224.3113.97
May 253.9618.73

The new surfaces split sources harder than the old pair

Classic ChatGPT logoChatGPT and Google logoGoogle AI Overviews shared 10 of their top 25 sources in the earlier study of matched prompts. ChatGPT logoChatGPT Search and Google logoGoogle AI Mode share only 6. Brand overlap moved the other way, rising from 16.4% on the older pair to 26.9% here.

  • Classic ChatGPT logoChatGPT + AI Overviews10 shared
  • ChatGPT logoChatGPT Search + AI Mode6 shared
Top-25 source overlap by engine pair.

The operating rule: split source work by surface

If you need Google logoGoogle AI Mode coverage, YouTube logoYouTube, social platforms, and broad web aggregation matter more than the ChatGPT logoChatGPT Search evidence mix suggests. If you need ChatGPT logoChatGPT Search coverage, vendor pages, specialist editorial, developer sources, and academic sources matter more.

A single AI-search checklist is too coarse. Track the sources separately from the brands, and measure each surface on the sources it actually reads.

Get the data

Dataset CSVHeadline figures behind the source, brand, and citation-count comparisons.

Sources

  1. Companion analysis: ChatGPT Search vs Google AI Mode sources
  2. Semrush: most-cited domains in AI search
  3. Profound: AI platform citation patterns

Related research

When ChatGPT and Google recommend the same top brand, do they cite the same page?
Almost never. When both engines chose and cited the same top brand, they shared a cited page for it in 309 of 8,612 matched answer pairs, or 3.59%.
ChatGPT vs Google: same winner, different shortlist
The popular line is that AI engines disagree about who wins. They don't. Across 1,655 buyer categories ChatGPT and Google AI Overviews pick the same #1 brand 93% of the time. What they disagree about is everyone else on the list.
AI citation volatility by industry: one-shot checks miss the signal
Ask the same AI question again and the cited sources usually change. Across repeated runs, ChatGPT answers shared only about 21% of their cited sources, and every industry showed high churn.
Which domains AI cites most in its answers
A source-domain cut of AI answers: which domains supply the evidence behind AI recommendations, and why community platforms are not a side channel.
Do AI-cited pages actually mention the brand?
Not always. In 80,623 of 209,116 reviewed citations, or 38.55%, the cited page did not mention the brand attached to the citation.
Does AI make brands sound better than cited sources?
AI often does. Across 512,650 brand-citation tone pairs, positive shifts outnumbered negative shifts 207,858 to 41,615, or 4.99 to 1.
How often do AI citations lead to broken pages?
Rarely. Of 152,313 AI-cited pages checked at or after their latest recorded citation, 1,015, or 0.67%, returned 404 or 410.
Does AI change its mind when comparing brands?
Sometimes. On the same prompt and criterion, one AI engine picked a different brand winner in 383 of 2,438 repeated matchups, or 15.71%.
How often do AI-cited pages block access?
Rarely. Of 154,986 pages checked after their latest recorded citation, 2,435, or 1.57%, returned 401, 403, or 429.
How often does AI cite a brand's own website?
Often. Of 1,060,513 page citations tied to ranked brand recommendations, 288,629, or 27.22%, pointed to the recommended brand's own website.
Is Reddit's share of AI citations growing?
No, not in this observed cut. Reddit's share fell from 4.88% to 3.82% on the same 15,769 prompts while YouTube rose from 3.86% to 5.18%.
Do AI engines cite their own websites?
Google's does, on most answers. Google AI Mode cited a Google-owned page on 15,833 of 18,304 matched prompts, or 86.50%. ChatGPT Search cited one on 2,014, or 11.00%.
Do ChatGPT and Google cite the same page?
Usually not. When both engines cited the same website for the same question, they landed on at least one identical page in 19,323 of 53,032 cases, or 36.44%.
How often does AI cite a competitor's website?
About one in twenty. Of 136,478 reviewed links between cited pages and recommended brands, 6,965, or 5.10%, came from a different brand in the same market.
How often does AI cite PDFs?
Rarely. Of 3,954,548 citations in ChatGPT Search and Google AI Mode answers, 16,075, or 0.41%, pointed to identifiable PDFs.
Does AI cite sources for its top brand recommendation?
Only about half. Of 215,349 top-ranked brand recommendations, 113,609, or 52.76%, had at least one citation attached to that brand in the same AI answer.
Does AI cite the same website more than once?
In nearly six in ten answers with at least two cited pages. AI cited multiple pages from the same website in 251,446 of 423,921 such answers, or 59.31%.
Do ChatGPT and Google recommend the same top brand?
Only about one in three times. ChatGPT Search and Google AI Mode chose the same top brand in 29,616 of 81,800 matched same-prompt answer pairs, or 36.21%.

About this research

Dimitry Apollonsky

Founder, Parse

I built Parse to track where AI answers really come from: the sources they cite and the brands they name. DM me on LinkedIn to talk shop.

See which AI search surface is reading your category.

Run a free check against live AI answers — no account needed.

Parse

Parse indexes AI recommendations so brands know where they stand.

Products

  • Brands
  • Markets
  • Integrations
  • Work with us
  • Pricing
  • MCP

Resources

  • Research
  • Methodology
  • Blog

© 2026 Parse. All rights reserved.

LegalPrivacy PolicyTerms of Service