Do ChatGPT, Perplexity and Claude Agree? 90 AI Answers

Last updated: November 2026

ChatGPT, Perplexity and Claude named the same top brand for only 7 of 30 buying questions (23%) when we asked each engine the same unbranded question on November 4, 2026. Their top five lists overlapped by less than half on average, and 46% of the brands named were recommended by just one engine.

That's 90 answers: 30 questions across 10 categories, asked once on each engine with web search on and the location set to the US. Being recommended by one AI engine tells you very little about the other two.

Key takeaways🔗

• All three engines picked the same #1 brand on 7 of 30 questions (23%). Exactly two agreed on 16 questions (53%), and on 7 questions (23%) each engine led with a different brand.

• Any two engines shared a top pick on 37% to 43% of questions. ChatGPT and Perplexity matched on 11 of 30, ChatGPT and Claude on 13, and Perplexity and Claude on 13.

• Top five lists overlapped by 0.48 on average (Jaccard similarity, where 1.0 means identical lists). On average only 2.2 brands appeared in all three top fives.

• Almost half the brands came from a single engine. Across the 30 questions, 114 of 247 brand appearances (46%) were named by only one of the three engines.

• Agreement depends on the category. Accounting software for freelancers had the most overlap (0.78) and mattresses the least (0.23).

How did we test ChatGPT, Perplexity and Claude?🔗

We asked the same 30 neutral buying questions to ChatGPT, Perplexity and Claude through their APIs, using the same engine settings TrendlyAI uses for its tracking. Each question was asked once per engine, on November 4, 2026, with web search enabled and a US location.

I wrote the questions the way a US buyer would type them, without naming any brand: "What is the best CRM for a small business with 5 to 10 employees?", "Which VPN would you recommend for someone who works on public Wi-Fi a lot?", and so on. There were three questions in each of 10 categories:

• CRM for small business

• Project management software

• Email marketing

• Accounting software for freelancers

• Running shoes

• Mattresses

• Password managers

• Website builders

• VPNs

• Online payroll

We then pulled out the ordered list of brands each answer recommended and normalized the names, so "HubSpot CRM" and "HubSpot" count as the same brand and "Brooks Ghost 16" counts as Brooks. Extraction was done by a model and checked by hand on 10 answers. One error turned up (an answer that named HubSpot and Salesforce only as integrations), so we tightened the rules and reran every extraction. All 90 calls succeeded.

How often do the engines agree on the top brand?🔗

All three engines put the same brand first on 7 of the 30 questions, or 23%. The most common outcome was partial agreement, where two engines shared a top pick and the third went elsewhere (16 questions, 53%). On the remaining 7 questions, no two engines agreed.

Looked at in pairs, the rates are similar. ChatGPT and Perplexity shared a top pick on 11 of 30 questions (37%). ChatGPT and Claude did on 13 (43%), and so did Perplexity and Claude.

The CRM questions show the pattern clearly. ChatGPT and Claude put HubSpot first on all three CRM questions. Perplexity didn't lead with HubSpot once: it opened with Pipedrive, Zoho and Less Annoying CRM. HubSpot still appeared in every Perplexity CRM answer, just lower down.

Do the engines at least recommend the same set of brands?🔗

Partly. Each engine's top five overlapped with another engine's top five by 0.48 on average, measured with Jaccard similarity (shared brands divided by all brands across both lists). In plain terms, two engines' shortlists for the same question had a little under half their brands in common.

Across the three engines together, the picture is less tidy:

• 2.2 brands on average appeared in all three top five lists.

• 8.2 different brands were named per question, on average, once you combine all three answers.

• 127 distinct brands were named across the whole study.

• 46% of brand appearances (114 of 247) came from only one engine.

The most extreme case was "What is the most affordable email marketing software for a small business?" The three answers named 13 different brands between them and led with three different ones: MailerLite (ChatGPT), Zoho (Perplexity) and Constant Contact (Claude). Only one brand, Brevo, appeared in all three top fives.

Which categories had the most and least agreement?🔗

Agreement varied a lot by category. Software categories with a few well-known leaders, like freelancer accounting and VPNs, had the most overlap. Mattresses, email marketing and website builders had the least.

With three questions per category, these are directional rankings rather than firm ones. A single question can move a category up or down the table.

A few category results stand out:

• Accounting: the same five names (FreshBooks, QuickBooks, Wave, Xero, Zoho) kept coming up everywhere, but no question had all three engines leading with the same one.

• Payroll: ChatGPT put Gusto first on all three payroll questions. Perplexity and Claude each put OnPay first on the multi-state hourly question.

• Mattresses: all three engines named Helix first for side sleepers. For lower back pain, ChatGPT led with Saatva, Perplexity with Helix, and Claude named no brands at all.

How many brands and sources does each engine give?🔗

ChatGPT named the most brands per answer, and Perplexity cited the most sources by a wide margin. ChatGPT also searched the web far less often than the other two.

Search behavior seems to matter. On the 11 questions where ChatGPT searched, its top five overlapped with Perplexity's by 0.64 on average. On the 19 where it answered from memory, the overlap was 0.46. That's a small sample, but it suggests some of the disagreement comes from one engine reading today's web while another relies on its training data.

Claude declined to name any brand in 2 of its 30 answers (beginner running shoes and mattresses for lower back pain). In both it didn't search and gave general buying advice instead. Neither of the other engines did this.

Note that the source counts aren't strictly like for like. Perplexity returns every source it used, while ChatGPT and Claude only count the links they cite inline.

What does this mean for brands?🔗

If you only check ChatGPT, you are seeing roughly one third of the picture. In this study, ChatGPT's top pick was missing entirely from Perplexity's answer on 5 of 30 questions and from Claude's on 4 of the 28 questions where Claude named brands. A brand can lead on one engine and not appear on another for the same question.

Here is what I'd take from the data:

1. Check every engine separately. A good result in ChatGPT doesn't carry over to Perplexity or Claude. Track your brand on each engine you care about, using the questions your buyers actually ask. Our guide to tracking brand mentions across ChatGPT, Perplexity and Google covers the setup.

2. Measure position, not just presence. Plenty of brands showed up on all three engines but in very different positions. HubSpot appeared in every CRM answer yet was never Perplexity's first pick. Share of voice and rank both matter.

3. Expect each engine to need its own work. Perplexity always searches and cites around 10 sources, so recent, well-cited pages matter there. ChatGPT and Claude sometimes answer without searching, so long-standing reputation and coverage matter too. See how to get cited by Perplexity, how to get mentioned in ChatGPT and how to get recommended by Claude.

4. Measure more than once. AI answers change from run to run. One check is a snapshot, and a trend over weeks is what tells you whether something is working. We are doing this with our own brand in public, starting from a Week 0 baseline.

This lines up with earlier work on sources. SE Ranking's April 2025 comparison found that the highest domain overlap between any two AI engines was 25.19% (Perplexity vs. ChatGPT). We looked at brands rather than domains, and found the same kind of split.

What are the limits of this study?🔗

This is a small, single-day snapshot, and it should be read as one. The main limits:

• One answer per engine per question. AI answers vary between runs, so the same test tomorrow would produce different lists. With 30 questions, each question moves a percentage by about 3 points.

• API models, not the consumer apps. We used gpt-4o-mini, sonar and claude-haiku-4-5 through their APIs. The ChatGPT, Perplexity and Claude apps people use day to day run different models, with their own system prompts, memory and personalization.

• The models chose whether to search. ChatGPT searched on 11 of 30 questions and Claude on 23. Some disagreement comes from that choice rather than from different opinions about the same sources.

• Brand extraction was done by a model. We checked 10 answers by hand and fixed one error, but some small mistakes may remain. We compared at the brand level, so two different Brooks shoes count as agreement.

• US, English, consumer and small business categories only. Results may differ in other countries, other languages and enterprise B2B categories.

Frequently asked questions🔗

Do ChatGPT, Perplexity and Claude recommend the same brands?🔗

Only partly. In our test of 30 unbranded buying questions on November 4, 2026, all three engines named the same top brand 23% of the time. Their top five lists overlapped by 0.48 on average (Jaccard similarity), and 46% of the brands named came from only one engine.

Which AI engines agree with each other most?🔗

The differences between pairs were small. ChatGPT and Perplexity had the highest top five overlap (0.53), while Perplexity and Claude had the lowest (0.43). On the top pick alone, ChatGPT and Claude matched on 13 of 30 questions, Perplexity and Claude on 13, and ChatGPT and Perplexity on 11.

Why do AI engines recommend different brands?🔗

They search differently and draw on different sources. Perplexity searched on every question and cited 10 sources per answer. ChatGPT (gpt-4o-mini) searched on 11 of 30 questions and Claude (claude-haiku-4-5) on 23. When ChatGPT searched, its shortlist overlapped more with Perplexity's (0.64 vs. 0.46), which suggests search behavior explains part of the gap.

If my brand shows up in ChatGPT, will it show up in Perplexity and Claude?🔗

Not necessarily. In this study, ChatGPT's top recommended brand didn't appear at all in Perplexity's answer to the same question 5 times out of 30, and was missing from Claude's answer 4 times out of 28. You need to check each engine separately.

How often should I check my AI visibility?🔗

Regularly, because answers change between runs and over time. A single check shows one moment, so tracking the same questions on each engine weekly or monthly gives you a trend you can act on.

Can I download the data?🔗

Yes. The full dataset is free to use with a link back to this page:

• Per-question results (CSV): top pick per engine, agreement and overlap for each of the 30 questions

• Results by category (CSV)

• Brand frequency (CSV): every brand named and by which engines

• Every answer's extracted brands (CSV)

• Methodology notes

Related reading🔗

• How to Track Brand Mentions in ChatGPT, Perplexity and Google

• What Is AI Share of Voice?

• Most-Cited Domains in ChatGPT

Want to know which engines recommend your brand and which don't? Run a free AI visibility check on ChatGPT and Perplexity. Paid plans track ChatGPT, Perplexity, Claude and Google AI Overviews every week. For the full playbook, see our GEO hub.