Two instruments. One dated before and after.
Instly Scout measures whether AI engines name your business when buyers ask, finds out why they do not, does the work, and measures again with the same instrument. This page is the whole method, with nothing left to trust us on: what each instrument checks, how it is scored, what a reading can and cannot tell you, and what happens at each step.
Instrument 1: website check
Thirteen weighted technical checks on your homepage, over plain HTTP, no browser. Deterministic: the same page gives the same score. Free, instant, no signup. Every failing check records what it saw: status codes, URLs tried.
Instrument 2: AI answer check
Five questions your buyers actually ask, put to ChatGPT, Gemini, Perplexity and Claude with web access on. For each answer we record which brands are named, in what order, and which pages the engine read. Reported as a count: named in X of 20.
The two numbers are never blended. A site can score 100 on the first and 0 of 20 on the second; ours did, on 5 Oct and 29 Sep 2026. Passing the technical checks is the floor. Being named is the goal.
What the 13 checks are and what each is worth.
148 points available, scaled to 100. A pass scores the full weight, a warn scores half, a fail scores nothing. Two further findings carry no points but are reported: brand not in the opening text, and sitemap pages with no internal link.
| Check | Weight | Pass means |
|---|---|---|
| Indexing allowed | 20 | robots and meta directives let search and AI crawlers index the page |
| JSON-LD schema | 18 | structured data an engine can read without guessing |
| llms.txt | 15 | a plain map of the site for AI engines |
| AI-crawler access | 15 | GPTBot, ClaudeBot, PerplexityBot, Google-Extended and the user agents all get a 200 |
| Non-browser client access | 12 | the page serves to a plain HTTP client, not only to a browser |
| Sitemap | 10 | present and valid, checked at every usual path |
| Canonical URL | 10 | one canonical, pointing at itself |
| Meta title | 10 | present and within the length engines show |
| Meta description | 10 | present and within length |
| H1 | 8 | one H1 with text |
| Bare domain | 8 | the bare domain resolves to the site |
| Image alt text | 7 | images carry descriptions |
| HTTPS | 5 | served over HTTPS |
A homepage that refuses the checker (HTTP 403) is recorded as inconclusive, never as a failure, and is not scored. A 429 or a timeout is the same. Absent is reported as absent only when every usual path was tried and what came back is recorded.
How the questions are set, and why they never change.
Setting the five questions
The questions are written from your own positioning, the way your buyers phrase them, and agreed with you before the first check. Not generic prompts like "best CRM": those tell you nothing. They stay fixed so that later checks are comparable.
What one reading can tell you
One mention is one twentieth of a check, so a single reading can move five points with nothing changed. We report counts, never percentages, and call a trend only on repeated movement. Counts of 0 to 3 are within noise. A reading says where you are on that day, with those questions, on those engines; it does not say why, until the pages the engines read are looked at.
Where the Scout report fits, and what its score is.
The Chrome extension and the USD 49.99 report
The extension scans up to eight pages of your site in the browser and produces the Scout report: the technical checks plus content checks (one clear heading, FAQ or answer blocks, review or case-study markup, a clear call to action), an executive summary and a prioritised fix list. The free scan shows the headline; the full report is unlocked for USD 49.99 and stays available under Sign in, My reports.
How to read its score
The report's 0 to 100 is an assessment across those pages against a written rubric, produced with a model. It is useful for deciding what to fix first. It is not the measurement we prove work with, because it is not reproducible to the point. Proof always comes from the two instruments above, which is why every number on this site says which instrument it came from and when.
What happens at each step.
- Free website check. Run it on the homepage. If it fails checks, those are fixable in a day and we will tell you how; if it passes, the technical side is not your problem.
- Call, twenty minutes. We agree the five buyer questions, who approves changes on your site, what platform it runs on and who else touches it. No card. Terms are agreed in writing before anything starts; nothing begins without them.
- Baseline. Both instruments run and you get the dated reading before any work. Without a baseline there is no evidence the work did anything, so we do not skip it.
- Diagnosis. Why you are not named, from what the engines actually read: which pages, which competitors, which questions. Stated as proven, or stated as unknown with the one check that would settle it. We do not charge for fixing something we cannot name.
- Work. Content written and blockers fixed. Every change is drafted by us and published only when a named person at your company says yes. Your site is production: only the agreed changes, never a platform, plugin or theme update.
- Re-measure and publish. Same five questions, same four engines, no earlier than four weeks after the work, because engines re-read on their own schedule. The before and after goes on your report page, dated, whichever way it lands. Our own case study shows a lost answer alongside the gained ones.
One sentence each.
Program
USD 500 a month, three-month term, no setup fee, additional sites USD 200 a month. If your named-in-AI count has not moved by the end of month three, month three is refunded and you keep everything produced.
Who runs it
One person: Mo Buglow, founder of Instly Technologies, Bangkok. The instruments on this site were built to stop our own platform reporting things that were not true, after it was found doing exactly that in August 2026. The scores we publish for our own sites are the real ones, including the zero. About.
Worked examples of everything above: the IPMIcompare case study (3 of 20 to 5 of 20, with every answer), the 20-vendor technical benchmark, the league table of who AI names, and the head-to-head comparisons.