One scan, in full

EXAMPLE DATA

Northlight · example.com · Manchester · ran 1 Sep 2026, 06:12 UTC

Ten prompts, two assistants, three runs each — 60 runs. Forty-eight came back. What the other twelve would have said is not knowable, so this page does not guess.

Every figure on this page is fabricated. A stored scan holds a real customer's brand, their competitors and where they are losing; it is confidential and never appears on a page in build/.

7 of 8

Prompts that named you

cited OR mentioned, in at least one run

denominator: prompts where an engine answered

3 of 8

Prompts that cited you

example.com was fetched and linked

denominator: prompts where an engine answered

5 of 8

Prompts that mentioned you

named from model memory, nothing fetched

denominator: prompts where an engine answered

4 checks

Not measured

of 20 prompt × engine checks — excluded from all three denominators

never counted as a miss

Why 3 + 5 does not make 7

Because a prompt can be both. On “top review management providers” ChatGPT cited your page and Gemini named you without fetching anything — that prompt is in the cited count and in the mentioned count. The two are different evidence and are kept apart on purpose: a citation proves an assistant went and read your page, a mention only proves it had heard of you. One combined “visibility” number would claim the first for the second, which is the single most common way this measurement is faked. There is no such number anywhere in this product.

Every prompt

Northlight · example.com · 1 Sep 2026 · 3 runs per prompt per engine
Prompt	ChatGPT	Gemini	Named instead
Best review software for contractors	citedcited in 3 of 3 runs	citedcited in 2 of 3 runs	—
Top review management providers	citedcited in 2 of 3 runs	mentionednamed in 3 of 3 runs, never fetched	competitor-one.com
Review software for home services companies	citedcited in 2 of 3 runs	missingnamed in 0 of 3 runs	competitor-one.com, competitor-two.com
Northlight vs competitor-one	mentionednamed in 3 of 3 runs, never fetched	mentionednamed in 2 of 3 runs, never fetched	competitor-one.com
Best tools for collecting customer reviews	mentionednamed in 2 of 3 runs, never fetched	mentionednamed in 3 of 3 runs, never fetched	competitor-one.com
How do contractors get more Google reviews?	mentionednamed in 1 of 3 runs, never fetched	mentionednamed in 2 of 3 runs, never fetched	competitor-three.com
Cheapest review management software	mentionednamed in 2 of 3 runs, never fetched	missingnamed in 0 of 3 runs	competitor-two.com
How much does review software cost?	missingnamed in 0 of 3 runs	missingnamed in 0 of 3 runs	competitor-two.com, competitor-one.com
Review software with QuickBooks integration	not measured— why rate_limited — the provider returned 429 on all 3 runs. No answer came back, so there is nothing to judge.	not measured— why upstream_error — the connection to the provider failed on all 3 runs.	—no answers, so no competitors could be read off them
Best review platform for small trades businesses	not measured— why rate_limited — the provider returned 429 on all 3 runs.	not measured— why upstream_error — the connection to the provider failed on all 3 runs.	—no answers, so no competitors could be read off them

“Named instead” is derived from the answers — domains cited where you were not, with your own domain and the usual aggregators removed. It is not a competitor list you or we curated, and a name appearing there is not a judgement about that business.

What did not run, and why

Four of the twenty prompt × engine checks got no answer at all — twelve of the sixty runs. Those four are not in any count on this page, in either direction. They are not misses, they are not zeros, and they are not quietly folded into the runs that did come back.

ChatGPT — rate_limited: 429 on 6 runs (prompts 9 and 10) Gemini — upstream_error: connection failed on 6 runs (prompts 9 and 10)

The next weekly scan retries them. If a check keeps failing, that is a thing to fix on our side, not a result about your brand.

How this scan was run

engines ChatGPT, Gemini — both answeredPerplexity is not on any plan yet: its per-call cost has never been measured against a live key, and putting an unmeasured engine behind a plan commits to a bill nobody has seen.

runs per prompt 3, per engineAI answers vary run to run, so every state on this page is k of n over three runs, not a single check.

model stamp gpt-4.1-2025-04-14 · gemini-2.5-flash Reported by the provider on 15 of the 16 measured cells. One cell is mixed — the provider rolled its build mid-scan, so that cell records both versions and names neither as the model. The 4 unmeasured cells are not reported: no answer came back to stamp.

scan id 3f9a1c0e-example-only-not-a-real-id

sharing Not sharedThis record is yours. It reaches a public page, a case study or an aggregate only if you tick a separate box saying so — never as part of accepting terms, and never for training a model.

Back to history Share of voice for this scan