Understanding and improving your visibility on AI assistants - practical guides, methods and benchmarks.
Two AI engines barely consult the same pages, yet agree far more on the companies they recommend. What the published research shows, and what it changes for your visibility.
Same trade, different country: one French web agency in four is named by AI. In the United States, three out of eighty-seven. A big market erases.
544 websites visited across twelve countries. 278 publish a readable address, 96 do not respond. In Japan, four businesses in five publish none.
Three identical runs, five sectors, two engines: two cited names out of three appear only once. And a citation's value depends on the engine - one in two holds on Perplexity, one in twenty on Gemini.
24 August: assistants do not read the same pages in every language. Today: they do not read the same NUMBER of them. In Busan, one domain is cited for all eleven trades.
Five runs of the same question: the brand named first stays the same 81 % of the time. The problem is not stability - it is that there is no single first place.
We sell repeated measurement. So we tested whether the list drifts from one day to the next. It barely does - and for one engine, not at all.
« The best X » and « which one for my 10-person company » overlap by only 31 %, against 77 % when the exact same question is asked twice.
We read the pages the AI says it consulted and looked for the recommended brand in them. 128 brands out of 128, with a median of six pages citing each one.
Yesterday: the recommended brands change with the language. Today: the pages read to pick them. 90 % of the 652 cited domains appear in ONE language only.
180 answers, 10 languages, 3 AI assistants. Overlap with English falls to 43% for online payments and 52% for web hosting - and stays intact for running shoes.
48 answers, 8 communes, from Paris to a sub-prefecture of 10,201 people. Gemini and Perplexity name real bakeries everywhere - with the address. ChatGPT answers without a question only in Paris.
3,780 real answers, 18 trades, three countries. ChatGPT declines to name a business in 43% of cases - 92% for solicitors, 91% for plumbers, 0% for bakers. Gemini: 1%. Perplexity: 0%.
18 conversations, three countries, two turns each. At the moment the assistant has to pick one, a third of the time it names a business that was never on its own list.
4,092 answers, three countries, 39 cities. ChatGPT names businesses in almost every answer - and in 100% of them it gives no link, no phone number and no address.
Measured 19 August 2026 across five markets. Of 22 businesses that Google's first page surfaces, Gemini names 8, Perplexity 3, ChatGPT none. And in Manchester, none of the three.
Yesterday we published a fix we refused to ship. One distinction was missing. Measured on the same 72 answers: 26 businesses found out of 28, against 23 - and none lost.
Measured across 72 answers: 454 extracted names for 424 real businesses. The obvious fix absorbed 61 duplicates - and made 72 real shops disappear. We did not ship it.
Measured 17 August 2026 across six markets. In Japanese and Korean, 100% of the names cited are in the local script - and our extractor required a Latin capital.
Read from our own Search Console on 16 August 2026. Impressions went up tenfold in six days - and almost all of them came from people looking for something other than us.
Measured on 15 August 2026 across six cities, in French, Portuguese and English. Re-asking in the same language returns 61% of the names; asking in another language, 20%.
Measured on 13 and 14 August 2026 across six French cities. Perplexity keeps half its names from one day to the next, Gemini a third - and ChatGPT still names no local business.
Measured on 13 August 2026 across six French cities. Of 42 businesses cited, none is named by all three assistants - and one of them names none.
Measured on 12 August 2026 across four trades and four cities. The engine reads thirteen pages per answer - and almost never the site of the business it recommends.
Measured on 11 August 2026 in two French cities. No business appears across all five phrasings - and on one of them, the AI names nobody at all.
Measured this morning across four trades. The result forced us to fix our own score - an engine that did not search no longer votes « absent ».
Measured 9 August 2026 on three very common French business names. With no city in the question, one engine elects a single establishment - and gives its address.
Re-measured this morning: of 81 names AI assistants cited on 6 August, only 15 reappear on the 8th. What holds for 48 hours, what moves, and why.
On 10 July 2025 Meta opened Instagram to search engines. Thirteen months on, it is the top source AI assistants cite in Brazil. A three-market investigation.
Measured 6 August 2026 on plumbers in two French cities: one refusal, one from-memory list, one directory roundup. Three AIs, three answer regimes.
We measured 24 Brazilian businesses in August 2026: 22 scored 0/100. One of them had won a client through AI the day before. Both things are true.
More customers now ask ChatGPT for “the best X near me.” Here's how to become part of the answer - and why it isn't SEO.
AEO (Answer Engine Optimization) isn't a new name for SEO. Here are the concrete differences and what they mean for your visibility.
Eight concrete actions, ranked by effort, to get ChatGPT, Gemini and Perplexity recommending you. Do them this month.