AI visibility readings: who the engines name.
These are Instly's own websites, measured by our own platform, with the date on every number. On 14 October 2026 we put the five questions a buyer of AI visibility tools asks to five AI engines: ChatGPT, Gemini, Claude, Perplexity and Google AI Mode. None of the 25 answers named instlyscout.com. The names the same answers gave instead: Profound, Otterly.ai, Scrunch and Peec AI, four answers each. The same site passes all twelve technical checks of the website check (100, 10 October 2026). Read technically, named nowhere: that gap is what this page tracks, on our own sites first. The table is the latest reading for each site. It redraws from the platform's public feed when the page opens; if the feed cannot be read, it shows the last published reading and says so.
Four sites, two instruments, every number dated.
Named in AI answers and the website check are two instruments. They are never added together, and a missing number means not measured, not zero.
| Site | Named in AI answers | Engines asked | Named instead | Website check (12 checks) |
|---|---|---|---|---|
| instlyscout.com AI visibility software | 0 of 25 14 Oct 2026 · 5 engines · naming rule v2 | ChatGPT, Gemini, Claude, Perplexity, Google AI Mode | Profound 4, Otterly.ai 4, Scrunch 4, Peec AI 4 | 100 10 Oct 2026 |
| ipmicompare.com IPMI comparison | 6 of 30 14 Oct 2026 · 6 engines · naming rule v2 | ChatGPT, Gemini, Claude, Perplexity, Google AI Mode, Google AI Overview | International Citizens Insurance 4, Pacific Prime 3, NowCompare 3, Cigna Global 3 | 100 10 Oct 2026 |
| instlytechnologies.com Insurance technology | 0 of 20 10 Oct 2026 · 4 engines · naming rule v1 | ChatGPT, Gemini, Claude, Perplexity | Cytora 3, Groundspeed Analytics 2, Sure 2, Cover Force 2 | 100 10 Oct 2026 |
| healthinomad.com Nomad health cover | 0 of 25 14 Oct 2026 · 5 engines · naming rule v2 | ChatGPT, Gemini, Claude, Perplexity, Google AI Mode | SafetyWing 4, Cigna Global 4, Genki 3, GeoBlue Xplorer 2 | 98 10 Oct 2026 |
Last published readings, 14 Oct 2026. The live numbers load from the platform when this page opens.
The feed behind this table is public: https://api.instlyscout.com/public/own-scores.json (JSON, our own sites only, cached ten minutes). The same table, with the Scout 13-check report beside it, is on the platform page.
The method, in plain words.
We put each site's buyer questions (five for every site here) to the engines we measure, one answer per question per engine. ChatGPT, Gemini, Claude and Perplexity are asked through each provider's API with web search on; Google AI Mode and Google AI Overview answers come from DataForSEO fetching Google's results page, for Thailand in English. The total is the number of answers checked, so five questions to five engines is 25. An answer that could not be checked is not counted, and the row says so.
An answer names a site only when its own words do: the configured name, an alias or the domain written in a sentence. A citation label, a link or a bare URL is not naming; a source label does not count. That is the rule we have used since 11 October 2026 (v2), and every reading carries the version it was judged by. A reading marked v1 was judged by the rule before it, which counted a source label as naming. We compare two readings only under the same rule and the same engines.
The names the same answers gave, with how many answers named each. This list comes from one model reading all of a site's answers together, so it is a tally of names, not a per-answer record; ties are shown in the order the platform gives them.
Twelve weighted technical checks on the homepage, 140 points scaled to 100, fetched the way AI crawlers fetch. A number is shown only when the check read the site itself and judged at least three quarters of the check weight; otherwise the row says why not, never a score. It is a different instrument from the 13-check Scout report (which adds the H1 check) and from the AI answer check.
AI engines do not answer the same way twice. A count of one to three can move by one between two runs on the same day: instlyscout.com read 1 of 19, then 0 of 19, both on 9 October 2026. A before and after is two readings taken under the same questions, engines and rule; one reading is not a trend.
The latest reading per site and per instrument, each with its own date. It does not carry the earlier series, so the swing from a baseline is written below by hand, from the platform's dated history, and will be updated when a new reading is published.
The series so far, dated.
From the platform's history, as of 14 October 2026. Readings under different rules or engine sets are listed, not compared.
instlyscout.com, named in AI answers
- 0 of 20 on every run from 27 July to 8 August 2026 (four engines, rule v1).
- 0 of 18 on 28 September and again on 29 September 2026 (four engines, rule v1).
- 1 of 19, then 0 of 19, both on 9 October 2026 (four engines, rule v1): the one answer did not repeat the same day.
- 0 of 18 on 13 October 2026, the first reading with Google AI Mode (five engines, rule v2).
- 0 of 25 on 14 October 2026 (five engines, rule v2): the reading in the table.
Every reading since 27 July 2026 has been zero except one answer on 9 October, which the same day's second run did not repeat.
ipmicompare.com, named in AI answers
- 3 of 20 on 28 September 2026, then 4 of 20 on 2 October 2026 after two fixes made on 28 September, under the naming rule we have used since 11 October 2026 (5 under the rule before it, which counted a source label as naming). A change of one answer is within noise.
- 6 of 30 on 14 October 2026 with six engines (rule v2). Not comparable to the 20-answer readings: the engine set changed.
The two fixes and what moved are in the case study.
The pages behind the 25 answers about instlyscout.com's questions.
From the platform's record of the 14 October 2026 run. These are the sources the five engines cited for the five buyer questions, by host, with how many answers cited each.
| Source | Answers citing it | Kind |
|---|---|---|
| geoptie.com | 6 | independent list (also a vendor) |
| tryprofound.com | 5 | competitor's own site |
| zapier.com | 5 | independent list |
| semrush.com | 5 | competitor's own site |
| reddit.com | 4 | social |
| otterly.ai | 4 | competitor's own site |
| blog.hubspot.com | 4 | independent list (also a vendor) |
| scrunch.com | 3 | competitor's own site |
Source mix across the 25 answers: independent articles and lists 74%, competitors' own sites 20%, social 6%, our own pages 0%. The platform classes geoptie.com and blog.hubspot.com as earned sources because the answers cited their round-up articles, though both companies also sell a tool.
What we are doing about it is on this site in the open: one page that answers each of the five questions directly (best AI visibility tools, checking AI citations, measuring AI search visibility, AEO tools, AI website audit), an honest comparison with the four tools named most, and this page. The next reading is published here whichever way it lands.
About these readings.
Why does instlyscout.com pass every technical check and still get named in 0 of 25 answers?
Because the two instruments measure different things. The website check asks whether an AI crawler can read the site: crawler access, schema, llms.txt, sitemap, titles. The AI answer check asks whether the engines name the site when a buyer asks a question. On 14 October 2026 the engines answered those questions from independent round-ups (geoptie.com, zapier.com, blog.hubspot.com), from the vendors own comparison pages and from Reddit threads, and instlyscout.com was on none of those pages. Technical readiness is the entry ticket; being written about is what gets a site named.
What counts as being named?
The answer must name the site in its own words: the name, an alias or the domain written in a sentence. A citation label, a footnote link or a bare URL under the answer does not count, and a source label does not count. This is the naming rule we have used since 11 October 2026, marked v2 on each reading; a reading marked v1 was judged by the rule before it, which counted a source label as naming.
Why do the totals differ between sites?
Each total is the number of answers checked: the site's questions multiplied by the engines asked, minus any answer that could not be checked. Five questions to five engines is 25; six engines is 30; a site last measured with four engines shows 20. The row says which engines were asked.
Can I quote these numbers?
Yes. Quote them with their date and their instrument, and link to this page or to the feed. They are readings of our own sites by our own platform. A reading of your site is its own run: the free scan covers the technical side at once, and the AI answer check is run on request.
The homepage scan is free and instant. The AI answer check for your buyers' questions is run on request.