On this page
The short version
We asked five AI systems the same 35 buying questions every day: ChatGPT, Gemini, Google AI Mode, Google AI Overviews and Perplexity. Then we looked at which business each one recommended most for each question.
- All five agreed on the same business once. One question out of 35, about 3%. Even that needed a tie. (The headline)
- When they all agreed, they mostly agreed nobody wins. All five agreed on 9 questions. On 8 of them, the "winner" was no business at all.
- A majority agreed more often. Three or more systems picked the same business on 14 questions (40%). (The full picture)
- The two Google systems agreed most (77%). ChatGPT and Perplexity agreed least (36%). (Which agree with which)
- "Visible in AI" isn't one thing. Each system needs its own check, and often its own fix. (What it means)
If you've ever asked "does ChatGPT recommend us?" and treated the answer as true for "AI" in general, this is for you.
The headline number
The short answer: five AI systems picked the same winner for 1 of 35 buying questions.
Here's how the 35 questions break down:
| What happened | Questions |
|---|---|
| All five picked the same business | 1 |
| All five agreed: no business wins | 8 |
| Three or more picked the same business (not all five) | 13 |
| Three or more agreed: no business wins (not all five) | 5 |
| No majority at all | 8 |
The one unanimous business was Intel Power Electric, for "24 hour electrician that covers Coquitlam and Port Moody". Even there, Perplexity had it tied with another firm.
Leave Perplexity out and the four older systems agreed on a business three times out of 35. Still under 10%.
I sell an AI visibility tracker, so weigh my interest in this result accordingly. The data is real, and one of the three businesses is our public demo.
How we measured it
The short answer: three real businesses, their real buying questions, five AI systems, every day.
Our tool, LLM Visibility Lens, asks each tracked question every morning in each AI system, without logging in, and records every business named in every answer. For this study we used the three accounts with the longest daily history:
- Kato Electrical, a Vancouver-area electrician: 16 questions, 5 September to 4 October 2026.
- An online mobility-equipment store (kept anonymous): 8 questions, same period.
- An online sod farm (kept anonymous): 11 questions, 16 September to 4 October.
That's about 3,000 answers in total. Perplexity was added to the tool on 23 September, so its answers cover about 10 days. The others cover the full period. Four newer accounts with less than a week of data are left out.
How "winner" and "agreed" are defined. A system's winner for a question is the business it named most often over the period; if more answers named no business than named any one business, the winner is "nobody", and names mentioned only once in the whole period are ignored. Systems agreed when their winners were the same; ties count, so if a system had two businesses level at the top, either one can match. Spelling variants of the same company are merged ("TDR Electric" and "TDR Electric Inc.", "Home Depot" and "The Home Depot").
Three of the sod farm's questions didn't have enough data in all four main systems, so they're left out. The full method behind the counts is on the how it works page.
Which AI systems agree with which
Some pairs agree far more than others, and the pattern makes sense.
The Google family sticks together. AI Mode and AI Overviews picked the same winner on 77% of questions. Gemini agreed with them 63% to 66% of the time. They're different products, but they draw on Google's index and Google's local data, so you'd expect overlap.
ChatGPT and Perplexity are the odd pair. They agreed on just 36% of questions, the lowest of any pair. Both are "chat" assistants, but they read very different sources. In our Perplexity vs ChatGPT comparison, Perplexity used about five times as many sources per answer and leaned on classic directories, while ChatGPT leaned on "best of" lists and official pages.
Even the closest pair disagrees often. 77% sounds high until you flip it: AI Mode and AI Overviews picked different winners on 8 of 35 questions. Ahrefs found the same thing at a much larger scale: the two cite different sources for the same query 87% of the time, as Search Engine Journal reported. We covered the Google pair in detail in Google AI Mode for local businesses.
Where they agreed
Agreement came in two very different forms: on nobody, and on a big name.
Agreeing on nobody. Eight of the nine unanimous questions had "nobody" as the winner. Most AI answers to them named no business at all. Examples:
- "How much does an electrical panel upgrade cost in Vancouver"
- "How much sod do I need and what will it cost for a quarter acre"
- Long, spoken-style questions like "What is the best lift recliner chair I can order online for my dad who has trouble getting up…"
These are cost questions, research questions and very specific requests. The AI systems agreed, in effect, that there was no business worth naming.
That's not a dead end. It's the most useful agreement in the data. A question every AI system leaves empty is an open door. We wrote about these "unclaimed" questions in We asked AI who to hire 80 times and about long, spoken questions in spoken questions.
Agreeing on a big name. When three or more systems agreed on an actual business, it was often a well-known one:
- Sod questions: Home Depot was the majority winner on four of them ("best place to buy online", "farm vs Home Depot quality", "fertilizer for new sod", "redo my front yard"). For a small online sod farm, the competition in AI answers isn't just other farms. It's a national retailer.
- Mobility store questions: Mobility Scooters Direct (heavy-duty scooters), Electric Wheelchairs USA (travel scooters), 1800Wheelchair and U.S. Medical Supplies each won a majority on at least one question.
- The electrician: Kato himself was a majority winner on "best electrician in Burnaby BC" (four of five) and "who is the best licensed electrician in Vancouver BC" (three of five). Intel Power won a majority on two Coquitlam questions, and TDR Electric on the EV charger question.
Want to see which questions in your market are unclaimed, and which already have a winner? The free question finder writes your customers' buying questions from your website.
Where they split
On 8 questions there was no majority at all. Each system had its own favourite.
The clearest example is the question you'd expect to have one obvious answer:
"Who is the best electrician in Vancouver BC?" ChatGPT's top picks were Whitley Electric and Arcwest Electric, tied. Gemini and AI Overviews went with Wicks Electric. AI Mode and Perplexity picked Kato.
So is Kato "the AI's" choice for best electrician in Vancouver? In two systems, yes. In three, no. A report that checked only ChatGPT would say he's invisible. One that checked only AI Mode would say he's winning. Both would be wrong about "AI".
The other split questions follow the same pattern. Emergency-electrician questions, "scooter with free shipping", "where can I buy St Augustine sod": three or four different winners across five systems.
You can open any of these questions in the live demo and read exactly what each system said, day by day.
Why they disagree
Three reasons, and none of them is a bug.
They read different sources. Google's systems lean on Google's own pages and local data. ChatGPT leans on rivals' websites, "best of" lists and official pages, as we found in ChatGPT citations. Perplexity casts a much wider net. Different reading lists produce different recommendations.
Each one is inconsistent with itself. Language models don't give identical answers to identical questions, as Thinking Machines Lab explained. SparkToro's study of AI recommendations found the same list of brands came back less than 1% of the time. If one system can't agree with itself day to day, five systems agreeing is a high bar. That's why we used the most-named business over weeks, not a single day. See why there's no rank #1 in ChatGPT.
Google decides when to show an AI Overview at all. For some local searches it shows a map instead, so the AI Overview "winner" comes from fewer answers. Our near me article covers when that happens. Google's own page on AI features and your website explains how it decides.
What it means for your business
"Are we visible in AI?" has five answers, not one.
- Check each AI system separately. Each one named the tracked business as its winner a different number of times. ChatGPT did it for 1 question. Gemini for 2, AI Mode for 3, AI Overviews for 2, and Perplexity for 5 of 33. A blended score would hide all of that.
- Expect the two Google systems to move together, and the chat assistants not to. Work on your Google presence (Business Profile, reviews, directories Google reads) tends to help AI Mode and AI Overviews at once. ChatGPT and Perplexity each need their own look. See how to rank in ChatGPT and how to rank in AI Overviews.
- Go after the questions everyone leaves empty. When all five agreed, it was usually on "nobody". Those are the cheapest wins: write the page that answers the question and names who does the work.
- Know who you're really up against. For an online sod farm, the AI competitor is Home Depot. For a local electrician, it differs by suburb and by AI system. Find out per question, not in general.
- Make crawlable your baseline. Let each system's crawler read you. Perplexity's crawler documentation and OpenAI's crawler documentation both explain what to allow.
How to track several AI systems
Ask the same questions in each system, on the same days, for a few weeks, and record who each one names. Then compare winners per question, not one overall number. Doing it by hand for five systems is slow, but it's possible. Our article on what one free check can and can't do explains how many answers you need.
LLM Visibility Lens does it automatically. It asks ChatGPT, Gemini, Google AI Mode and AI Overviews every morning, and Perplexity daily on the Pro and Agency plans. It shows which businesses each one names, and lets you read every answer.
Each month it turns that into an action plan: the questions no system has a winner for, the competitors each one prefers, and the sites to get listed on.
Agencies can send the monthly report under their own name, with every AI system reported separately. See the agency plan.
The plans start free. There's also a plain comparison with Semrush's AI visibility tool, the changelog, the FAQ page and the security page. If you'd like me to look at how the AI systems treat your business, get in touch.
Questions people ask
Do ChatGPT, Gemini and Perplexity recommend the same businesses?
Mostly not. Across 35 buying questions tracked daily, ChatGPT, Gemini, Google AI Mode, AI Overviews and Perplexity named the same top business for only one question. Three or more agreed on a business for 14 questions (40%), and 8 questions had no majority at all.
Which AI systems agree with each other most?
In our data, Google AI Mode and Google AI Overviews agreed most, picking the same winner on 77% of questions, and Gemini agreed with them 63% to 66% of the time. ChatGPT and Perplexity agreed least, on 36% of questions.
Why do different AI systems recommend different businesses?
They read different sources, they give slightly different answers each time even to the same question, and Google only shows an AI Overview for some searches. Google's systems lean on Google's own data, ChatGPT leans more on other websites and list sites, and Perplexity reads many more sources per answer.
If ChatGPT recommends my business, will other AI systems too?
Not necessarily. In our data, ChatGPT had the tracked business as its top pick for 1 question, while Perplexity had it for 5 and AI Mode for 3. Being recommended in one system says little about the others, so check each one separately.
What does it mean when AI systems agree that no business wins?
It means most of their answers to that question name no business at all, usually because it's a cost, research or very specific question. That's an opportunity: a clear page that answers the question and names who does the work has little competition.
How should I measure visibility across several AI systems?
Ask the same buying questions in each system on the same days for several weeks, and record which businesses each one names. Compare the results per question and per system rather than combining them into one score.
Harjeet Singh Bedi Founder, LLM Visibility Lens · SEO consultant, Bedi SEO
Harjeet has worked in SEO since 2010, first in agencies and now running Bedi SEO for local businesses, online stores and agencies. He built LLM Visibility Lens after clients kept asking whether they showed up in ChatGPT and other AI tools and nobody could give them an honest, measured answer. Every figure in this article comes from answers his tool collected.