AI search demand against AI citations, measured

This is reading one of a quarterly series, read 20 September 2026. The next reading is due 20 December 2026 and will be published beside this one. One reading carries no trend, so this page prints no change and no year on year figure.

Key facts

• AI searches a month: 2,047,996 across the 30 designer names within stocks, United States, read 20 September 2026
• Ten of those names: 1,494,666 of that total, 73.0 percent, and the ten are named in full below, read 20 September 2026
• Times within was cited as a source: 5, against TheRealReal 2,085 and Garmentory 65, counted where an assistant cited the domain as a source, ChatGPT, United States, read 20 September 2026
• The same domains counted the other way: within 5 and TheRealReal 4,607, counted where the domain was cited or retrieved and left unused, ChatGPT, United States, read 20 September 2026
• Canada, like for like: within 0 and SSENSE 1,860 on Google AI, counted where an assistant cited the domain as a source, because the vendor's panel holds no Canadian ChatGPT prompts at all, read 20 September 2026
• Keywords carrying AI volume: 624 of the top 1,000 in the United States and 588 of 1,000 in Canada, read 20 September 2026
• The widest gap: chanel, 114,934 AI searches in the United States against 0 Google volume in within's own baseline, read 20 September 2026
• Where ChatGPT data exists: 1 of the 94 locations the vendor covers, the United States, read 20 September 2026
• Boutiques on within: 2, both in Toronto, read 20 September 2026
• Readings published: 1, read 20 September 2026, with the next of this quarterly series due 20 December 2026

AI search demand against what within holds

One row per designer name in the reading, ordered by AI search volume. The volume column is monthly searches in the United States. The pieces column is what within's two Toronto boutiques had on the shelf when the index was read.

Jordan alone carries 520,265 of the 2,047,996. Ten of the names carry 1,494,666 between them, which is 73.0 percent of the total. The other 20 names carry 553,330.

Those ten are Jordan, Nike, Essentials, Louis Vuitton, Supreme, Chrome Hearts, Undercover, Rick Owens, Amiri and Kith. They are the names within's offsite pieces quote, and they are not the ten largest in this reading. The page names them rather than leave a reader to guess which ten a subtotal covers.

Undercover's volume was read and its piece count was not, so that cell is blank rather than filled. Nothing on this page is estimated. Piece counts move as pieces sell, which is why every one of them carries 20 September 2026.

Which retailers the answers cite, counted where an assistant cited the domain as a source

Nine domains in one call, ordered by the number of ChatGPT answers that named the domain among their sources. within is last with 5. The volume column is the monthly AI search demand on the questions those answers came from, added up: within's 206 is the 5 questions further down this page added together.

The same nine domains, counted where the domain was cited or retrieved and left unused

This is what the vendor answers when nobody sets the search_scope field, and it counts two things at once: an answer that cited the domain, and an answer that retrieved the page among its web results and quoted nothing from it. It is why TheRealReal is 4,607 here and 2,085 in the table above, the same domain on the same day. within is 5 in both, because every answer that pulled a within page quoted it. Quote either table, with the words above it.

The same nine domains on Google AI in Canada

The vendor's panel holds no Canadian ChatGPT prompts at all, so the like for like Canadian reading is Google AI: counted where an assistant cited the domain as a source, within is 0 and SSENSE is 1,860, read 20 September 2026.

A separate table because it is a separate measurement. DataForSEO returns no ChatGPT data for Canada at all, so these are Google AI figures and nothing here is a ChatGPT number. They are counted the way the second table above is, counted where the domain was cited or retrieved and left unused. Reading them beside the tables above is the point; adding them to one would be wrong.

Every answer that cited within

All 5 of them, which is the complete set the vendor holds rather than a sample. Each row is the question a reader asked, the page the answer linked, and where that link sat in the answer's source list. Every one of them cited within as a source, which is why within's count is 5 under both measurements above.

How much of the demand this channel holds

within's own keyword tracker, top 1,000 terms, sent to the AI volume endpoint once per country on 20 September 2026.

The two demand curves are not the same curve. These eight terms carry heavy AI volume and the Google column is within's own baseline for the same word, read 20 September 2026.

How this index was read

Search volume is DataForSEO's ai_optimization/ai_keyword_data/keywords_search_volume/live endpoint, United States, English, read on 20 September 2026, queried on the bare designer name with no modifier. Citation counts are DataForSEO's ai_optimization/llm_mentions/multi_target_metrics/live endpoint against ChatGPT answers, United States, English, read the same day, nine domains in a single call, once per scope. The answers behind within's own five came from ai_optimization/llm_mentions/search/live on the same day, model gpt-5-6.

The two citation tables are two different questions, and the page keeps them apart. Each domain the endpoint is given takes a search_scope field. Set to sources, a response counts only where the assistant cited that domain as a source, which is the headline table, read 20 September 2026 at 10:07 UTC. Left unset, the vendor's default is any: a response also counts where the assistant retrieved the page among its web results and quoted nothing from it, which is the second table, read the same day at 07:38 UTC.

The difference is not small. TheRealReal is 2,085 under sources and 4,607 under any, the same domain on the same day. within is 5 under both, because every answer that pulled a within page quoted it. Anyone quoting a number from this page should carry the words above the table it came from.

Piece counts are read from within's own A to Z designer index on 20 September 2026, which reads each boutique's own register. within recounts on every request, and the two boutiques sell over the counter all day. A piece count here is a reading at a moment rather than a standing figure.

Anyone with a DataForSEO account can reproduce the volume column and the citation column. The piece counts are ours.

The limitation that matters most: DataForSEO returns no ChatGPT data for Canada at all. Its own coverage call, read 20 September 2026, lists 94 locations. Exactly one carries the ChatGPT platform, the United States, over 164,118,449 English answers.

Canada carries 11,397,545 English answers on the Google platform alone. Every Canadian figure on this page is therefore Google AI, and it is labelled that way. The vendor's panel holds no Canadian ChatGPT prompts at all, so the like for like Canadian reading is Google AI: counted where an assistant cited the domain as a source, within is 0 and SSENSE is 1,860, read 20 September 2026.

This is one reading on one date, and it is reading 1 of a quarterly series. There is no earlier reading on the same basis to compare it to, so this page prints no trend, no change and no year on year figure. The next reading is due 20 December 2026 and will be published beside this one rather than in place of it.

What is not here: nothing is modelled, estimated or filled in. A figure that was not read is blank and says so, which is why Undercover's piece count is empty.

Citation counts are what one vendor's panel of answers recorded. They are not a census of every answer every assistant has given. A live sample of six buying questions on 20 September 2026 returned within 0 times.

Vendor coverage on 20 September 2026: 94 locations, of which 1 carries ChatGPT, the United States, over 164,118,449 English answers. Canada carries 11,397,545 English answers on Google alone.

What we think this means, which is an inference and not a measurement

The pages answer engines cite in this category are designer stockist pages and editorial. All 5 of within's citations are journal pages, read 20 September 2026. Not one is a directory page, a product page or a city page.

Nobody has measured whether article length, a table, or a date on a number changes whether a page is quoted. We are not going to claim they do.

Common questions

Citing these numbers

Anyone may quote these figures, in whole or in part, with attribution to within and a link to this page. They are published under a Creative Commons Attribution 4.0 licence, which asks one thing in return: say where they came from. The line to use is this one.

Source: within, within AI Visibility Index, read 20 September 2026, shopwithin.co/index/ai-visibility

The same numbers, the same method fields and the same read date are published as a machine readable document at /index/ai-visibility.json. For a figure this page does not carry, or a designer you would like counted, write to us.

What within is

2 boutiques sell through within, both of them in Toronto, read 20 September 2026. Between them they hold 128 designers with a live shelf. That is the whole of it, and it is why within sits last in both citation tables above.

• The Toronto Resale Price Index, our other data page
• Every designer stocked in Toronto
• The boutiques that sell through within