Findrix founder. AI search & brand visibility, distilled. Daily breakdowns of how AI picks what to recommend—for founders, marketers & the relentlessly curious.
Source: Search Engine Roundtable (Barry Schwartz), Nov 13, 2026. Glenn Gabe posted screenshots of a test: no AI Mode tab in the results tab bar, and no "Ask anything" box under an AI Overview.
https://t.co/8Q9RkPR45k
Barry wrote: "I cannot replicate this." The article also notes earlier reports of AI Mode buttons missing from the Google homepage.
Google is testing results with no AI Mode tab and no Ask anything box under AI Overviews.
I opened the Nov 13 writeup. One person posted screenshots. Barry Schwartz could not reproduce it.
I'd check both entry points today. I'd keep AI Mode and AI Overviews as two lines.
Tick one box on a Findrix Radar market page and most of ChatGPT's podiums turn into shared places.
The box is called "Only distinguishable places". With it on, Radar keeps a separate place only where it can tell two brands apart. In US marketing software, HubSpot and Salesforce end up sharing first and second.
It's there because a ranked list looks more certain than the answers behind it. A team chasing one place up is often chasing noise.
Across our 74 markets, 151 of the 222 top-3 places on ChatGPT overlap the next brand's range.
Four sites AI engines cite for local businesses that I rarely see on an SEO checklist:
- BBB. In ChatGPT's top 10 for three US trades markets: electrical, pool and spa, roofing.
- ProductReview. In ChatGPT's top 10 for five Australian local-service markets, from plumbing to conveyancing.
- LinkedIn. In ChatGPT's top 10 for five of our eight legal markets.
- Modernize. A contractor-matching site, and Gemini's most-cited source for US plumbing.
Each engine keeps its own off-site list, and none of these is your website. You can check all four this week.
On Gemini, https://t.co/bMBwC07L29 carries 33.1% of US plumbing citations.
Findrix Radar, AU dental, Oct 14, 2026, 400 questions. AI Mode: Pacific Smiles 57.5% (52.6–62.4), National Dental Care 12.5% (9.2–15.8). Gemini: 12.3% (9.0–15.5), overlaps #2. ChatGPT: not in top 20.
https://t.co/c84KzWWeoe
Australian dental is where I'd send anyone still arguing about AI visibility reports.
One engine or several? Several. Pacific Smiles Dental leads Google AI Mode, named in 57.5% of answers. On ChatGPT it isn't in the top 20 we publish.
First place or a range? A range. On Gemini it ranks first too, but inside a tie with the clinics behind it.
Same questions on every engine. Which argument would you add to this list?
Source: Christopher Jan Benitez, "Do AEO Best Practices Get You Cited by AI?", Nov 11, 2026. 51 US searches on Nov 10. Cited beat skipped 44% on AI Overviews, 41% ChatGPT, 55% Perplexity. One day, US only. He built the checker and sells AI visibility work.
https://t.co/97mxKsy4Ga
Christopher Jan Benitez compared pages Google, ChatGPT and Perplexity cited with pages they skipped, on 15 AEO checks.
Cited pages scored higher in 41 to 55% of pairs. A coin flip is 50%.
I'd keep the checklist as the floor. After an edit, do you re-run it or the questions?
Findrix Radar, October 2026: 39 Australian markets, each buyer question asked once per engine. Example, criminal defence law, top 10 sources per engine:
https://t.co/HROyKuRQlo
ChatGPT's favourite source for Australian service questions is often a government website.
In criminal defence, conveyancing, NDIS plan management and electrical services, the site it cites most is https://t.co/r84i7JTvAN or https://t.co/22ZXVXiwNo, ahead of every firm's own pages.
No firm can buy space there. What you can do is make your pages agree with it: the same terms, the same steps, the same eligibility rules, in the same words.
Gemini mostly looks elsewhere. In those same markets its top source is usually a provider's own site.
Across our Australian markets, a .gov.au site is ChatGPT's most-cited source in 15 of 39.
If you sell a regulated service in Australia, I'd read that government page before rewriting your homepage.
Findrix Radar, AU colour cosmetics & makeup, measured Oct 14, 2026, 800 questions, each asked once per engine. ChatGPT: Maybelline 52.4% (48.9–55.8), Maybelline New York 36.5% (33.2–39.8). Gemini 31.4% / 20.4%. AI Mode 25.4% / 16.8%. Names as the engines wrote them.
Maybelline shows up in ChatGPT's Australian makeup answers under two names. That's the first thing I'd check in any AI report.
Search the brand list for your name written another way. On our page, Maybelline and Maybelline New York are separate rows, as the engine wrote them.
"Maybelline New York" alone is in 36.5% of ChatGPT's answers. One answer can carry both, so the rows don't add up.
Count one spelling and you're measuring a smaller brand than buyers hear about.
Source: OpenAI Help Center, "Setting up your business on ChatGPT". For an online business, a mention counts when ChatGPT references the business's website. Naming the brand alone doesn't count. Insights cover the past 28 days, ChatGPT only.
https://t.co/Hd5aQYVSK7
Announcement: OpenAI, "Introducing Businesses in ChatGPT", November 10, 2026.
https://t.co/KlYuFBeL6M
ChatGPT's new mention count, for an online business, only includes answers that cite the site.
OpenAI's own help page: "The brand name alone doesn't count."
I'd treat that number as a citation count, not as how often the brand got named.
Findrix Radar, 400-question markets (51 of 52 use this split), questions written by Findrix, not buyer logs: general 6, use case 209, audience 20, selection criteria 120, problem to solve 45. AU mortgages shown, measured Oct 14, 2026, with examples per type:
https://t.co/ujMoh0Rec0
One check for anyone tracking AI answers on home loans: open your prompt list and count the questions that start with "best".
In our 400-question markets, only 6 are general, like "what are the best mortgage lenders in Australia?". About half, 209 of 400, are use cases, like finding a lender for an off-the-plan apartment purchase.
We wrote that set ourselves. Still, a score built only on "best X" measures the easiest slice.
What did your count come out to?
Test seen Nov 10 by Search Engine Roundtable, not announced by Google. It still pulls from the Business Profile; only the clickable parts are gone.
https://t.co/ymKbYlxvTC
Google is testing an AI Overviews local pack that keeps the business name and drops the buttons.
No call, no website, no directions, no reviews, no hours. Spotted by Joy Hawkins on mobile, replicated by Search Engine Roundtable.
I think you can be named and still lose the visit.
Nike is the activewear brand ChatGPT and Gemini name most in the US. On Google AI Mode, it's Lululemon.
Same buyer questions on each engine, and on all three the leader's range clears the next brand's.
AI Mode names Lululemon in 42.9% of answers. Nike ranks third there.
Which engine is your "we lead in AI" slide built on?