The short answerMostly not. Across 987 same-day answer sets from ChatGPT, Perplexity and Google AI Overviews on B2B agency questions (1 to 7 October 2026, Peec AI), 500 had all three naming an agency. In those 500, 71% of the agencies named appeared in only one engine's answer, and the three picked the same agency first 9.6% of the time.
| What we measured (1 to 7 Oct 2026) | Share | Raw count |
|---|---|---|
| Agencies named by only one of the three engines | 71.3% | 3,019 of 4,235 |
| Agencies named by all three engines | 7.8% | 332 of 4,235 |
| Answer sets with no agency common to all three | 54% | 270 of 500 |
| Same first-named agency on all three engines | 9.6% | 48 of 500 |
| Cited domains shared by all three engines (934 sets) | 0.7% | 245 of 32,804 |
If you track your AI visibility on one engine, you are reading one engine's shortlist. The other two are, most days, a different list.
How we ran the study
We track 150 buyer prompts for LoudFace in Peec AI, a tool that runs each prompt through ChatGPT, Perplexity and Google AI Overviews every day and records each answer, including its sources and the brands it names. 141 of those prompts are non-branded, by Peec's own tag: questions like "Best SEO agency for B2B SaaS companies" or "Which agencies do LLM optimization for B2B tech companies?".
We pulled every answer from 1 to 7 October 2026: 3,150 chats. Keeping only the non-branded prompts where all three engines answered on the same day left 987 answer sets of three answers each, 2,961 answers in total.
For each set we compared the agencies each engine named. "Named" here means one of the 53 brands Peec tracks in this project, almost all of them agencies. An agency outside that list does not count, so the study measures overlap among known names, not complete answer lists.
We used three measures:
- Single-engine share. Of every agency named in a set, how many appeared in only one of the three answers.
- Overlap. The agencies two engines both named, divided by the agencies either one named. 1.0 means identical lists. 0 means nothing in common.
- Top pick. Whether the first agency named was the same.
71% of named agencies show up on one engine only
In 500 of the 987 sets, all three engines named at least one tracked agency. Those 500 sets hold 4,235 agency appearances, counting each agency once per set:
| Named by | Agencies | Share |
|---|---|---|
| One engine only | 3,019 | 71.3% |
| Two engines | 884 | 20.9% |
| All three engines | 332 | 7.8% |
In 270 of the 500 sets (54%), no agency made all three lists. In the other 230 (46%), at least one did.
Here is one set, the prompt "Best SEO agency for B2B SaaS companies" on 1 October 2026:
| Engine | Agencies named |
|---|---|
| ChatGPT | Omniscient, SimpleTiger, Skale, First Page Sage |
| Perplexity | Omniscient, Directive Consulting, Siege Media |
| Google AI Overviews | Skale, Revv Growth, SimpleTiger, Siege Media, TripleDart |
Every pair of engines shares at least one agency. No agency is on all three lists. A buyer who asks one engine walks away with a different shortlist from a buyer who asks another, on the same day.
We ran the same count with LoudFace removed from every answer, in case our own brand was skewing it. The result did not move: 71.3% of agency appearances were single-engine (2,750 of 3,857), and 256 of 479 sets (53.4%) shared no agency across all three engines.
The engines rarely agree on the top pick
The first agency in an answer is the one a buyer reads first. The three engines put the same agency first in 48 of 500 sets: 9.6%.
Pair by pair, agreement is higher but still low:
| Engine pair | Same first-named agency | Rate |
|---|---|---|
| ChatGPT and Perplexity | 135 of 541 | 25.0% |
| ChatGPT and Google AI Overviews | 134 of 756 | 17.7% |
| Perplexity and Google AI Overviews | 97 of 519 | 18.7% |
Full-list overlap is just as low. ChatGPT and Perplexity share 0.225 on average, ChatGPT and Google AI Overviews 0.179, and Perplexity and Google AI Overviews 0.175. Roughly one agency in five appears on both lists.
Our numbers sit close to the only other same-day brand study we found. Vercite, an AI visibility tracker, reports that "When two engines answer the same prompt on the same day, they name the same brand first only 28% of the time." Its sample is 65,746 answers to 1,509 prompts across five engines, with a prompt mix that leans towards finance, real estate, hospitality, energy and business services. Ours is one B2B category in US English. Different samples, same direction, and in both, ChatGPT and Perplexity agree with each other most often.
Is it randomness? Each engine agrees with itself nearly twice as much
AI answers vary from run to run. SparkToro found "a <1 in 100 chance that ChatGPT or Google's AI, if asked 100X, will give you the same list of brands in any two responses." Even at settings meant to be deterministic, Atil et al. (arXiv, 2024) found accuracy variations of up to 15% across naturally occurring runs of large language models. So the fair question is whether the engines disagree with each other more than they disagree with themselves.
They do. We compared each engine's answer to the same prompt on consecutive days:
| Comparison | Average overlap | Answer pairs |
|---|---|---|
| ChatGPT vs ChatGPT, next day | 0.421 | 682 |
| Perplexity vs Perplexity, next day | 0.423 | 316 |
| Google AI Overviews vs itself, next day | 0.313 | 640 |
| Engine vs a different engine, same day | 0.175 to 0.225 | 519 to 756 |
An engine's answer today overlaps with its own answer yesterday nearly twice as much as with another engine's answer today. Some of the churn is noise. The rest of the gap comes from the engines.
Google AI Overviews is the least consistent with itself, at 0.313. If you only check it once a week, expect its list to look different each time you look.
They read different sources, too
We counted every domain each engine cited or linked in the 934 sets where all three engines cited at least one domain. Of 32,804 domain appearances, 30,490 (92.9%) came from only one engine. 245 (0.7%) were cited by all three.
Average overlap in cited domains is 0.032 between ChatGPT and Google AI Overviews, 0.051 between ChatGPT and Perplexity, and 0.060 between Perplexity and Google AI Overviews. The engines are not reading the same web and summarising it differently. They are reading different pages.
That matches an earlier check of our own answers: from 17 to 23 September 2026, 624 of 631 namings of LoudFace were in answers that read our site. If naming follows reading, engines that read different pages will name different agencies. The source split in this week's data is consistent with that pattern at category scale.
Ahrefs found the same split at the domain level, measured differently: "Only 14% of the top 50 mentioned sources were shared across ChatGPT, Perplexity, and Google AI". Its unit is each engine's top 50 domains across all prompts. Ours is per answer, which likely explains most of the gap.
Why is my brand in Google AI Overviews but not ChatGPT?
Because Google AI Overviews reads different sources. In our week of answers, 12.1% of the domains Google AI Overviews cited were LinkedIn, Reddit or YouTube (814 of 6,737): LinkedIn 5.7%, Reddit 3.5%, YouTube 2.9%. For Perplexity it was 4.0%, mostly LinkedIn. For ChatGPT it was 0.4% (92 of 20,942).
| Engine | Domains cited | LinkedIn, Reddit or YouTube |
|---|---|---|
| Google AI Overviews | 6,737 | 12.1% |
| Perplexity | 8,814 | 4.0% |
| ChatGPT | 20,942 | 0.4% |
The split is not new. In August, Google AI Overviews' top domains in our category already included youtube.com at 5.2%, linkedin.com at 4.2% and reddit.com at 2.3%, while listicles made up 81% of Perplexity's citations.
So a brand that shows up in Google AI Overviews and not in ChatGPT is likely to have its proof on LinkedIn posts, Reddit threads or YouTube videos that Google reads and ChatGPT barely cites. The reverse also happens: ChatGPT cites far more domains per answer, and a page Google AI Overviews does not cite can still be on ChatGPT's reading list.
Even the most-agreed agency is named by all three a third of the time
Within the 500 sets where all three engines named an agency, the one that most often made all three lists was Omniscient: 119 sets out of the 364 where any engine named it (32.7%). Next were iPullRank, 35 of 202 (17.3%), and First Page Sage, 31 of 254 (12.2%).
That is the ceiling. The best-known agencies in the category, the ones every engine knows, still miss at least one engine in most answers that name them at all.
At the prompt level, we looked at the 96 prompts where all three engines named an agency on at least three days of the week. 18 of them never had an agency common to all three on any day. 12 had one every such day.
The engines also differ in how often they name any agency. ChatGPT named a tracked agency in 854 of 987 answers (86.5%), Google AI Overviews in 822 (83.3%) and Perplexity in 572 (58.0%). When it names any, ChatGPT names 4.18 agencies on average, Perplexity 3.48 and Google AI Overviews 3.39.
How this compares with other overlap numbers
Published overlap figures disagree because they measure different things:
| Study | What it compares | Unit | What it found |
|---|---|---|---|
| This study (LoudFace, Oct 2026) | ChatGPT, Perplexity, Google AI Overviews, same prompt, same day | Agencies named | 7.8% named by all three; 9.6% same top pick |
| Vercite (2026) | Two engines, same prompt, same day | First-named brand | 28% same |
| SparkToro (2026) | One engine against itself, repeated runs | Full brand list | Under 1 in 100 identical |
| Ahrefs (2025) | Top 50 sources per engine, all prompts | Domains | 14% shared by all three |
| Ahrefs (2025) | AI citations vs Google's top 10 | URLs | 12% for ChatGPT, Gemini, Copilot |
| Semrush (2025) | Each engine vs Google's top 10 | Domains | Over 91% for Perplexity |
The Semrush and second Ahrefs figures measure each engine against Google's search results, not engines against each other, so a high number there does not mean the engines agree. One widely repeated figure, an 11% overlap between ChatGPT and Perplexity domains, has no method we could trace, so we left it out.
What this means for your AI visibility tracking
If your report shows one AI visibility number, it is hiding which engine you are losing.
Track each engine on its own line. A blended visibility number averages three different shortlists into one that no buyer ever sees. Peec defines visibility as the share of responses that mention your brand. Read it per engine, every time.
Read a week, not a day. A single run of one engine is one draw. Google AI Overviews overlapped with its own previous day by 0.313 on average, so one check tells you little. Use a fixed prompt set over at least seven days.
Get onto each engine's reading list. The engines name what they read, and they read different pages. Winning each engine means being on the pages it cites: lists and comparisons for Perplexity, where listicles made up 81% of citations in August, and the LinkedIn, Reddit and YouTube corpus Google AI Overviews leans on.
Expect to win one engine first. With 71% of named agencies appearing on one engine only, expect gains on one engine without a matching move on the others. Judge each engine against its own baseline.
If you want a first read on your own brand, our free AI visibility audit checks where you stand across ChatGPT, Claude, Gemini and Perplexity against your top competitors. For a do-it-yourself version, how to run a share-of-answer audit in 90 minutes is a manual playbook covering ChatGPT, Perplexity and Gemini.
For how the engines treat the consultant question specifically, see who AI names when you ask for an AEO consultant. For the sources they cite across the category, see our 90-day citation study, and for why naming follows reading, AI engines name brands they read.
Limits of this study
- Tracked agencies only. "Named" means one of the 53 brands Peec tracks in this project, almost all of them agencies. An agency outside that list does not count, so we measured overlap among known names, not complete answer lists.
- One week, one category. 1 to 7 October 2026, B2B SaaS SEO, AEO and growth-agency questions, US English. Other categories may agree more or less.
- One run per engine per day. Each prompt ran once a day on each engine. More runs would tighten the self-consistency numbers.
- Detection. We spot-checked five answers against their text. Every detected agency was in the answer, with one name written as an alias.
Measure each engine as its own shortlist, because that is how your buyers meet you.
