Answermend
example.com · owner Sign out

Brands / Northlight

Northlight

example.com · Manchester EXAMPLE DATA

Six scans, weekly. One of them measured nothing at all, and it is still in the record below — a scan that failed is not a scan that found nothing.

Every figure on this page is fabricated. A real scan record holds a customer's brand, their competitors and where they are losing, and never appears on a page like this one.

What changed

append-only · newest first

One entry per scan. Entries are added, never rewritten, so the record of a week we could not measure survives the week we could.

  • 1 SEP 2026 · 06:12 UTC

    Held a citation on three prompts for a second week

    Cited in 3 of 8 measured prompts, mentioned in 4 of 8. See this scan in full.

    4 of 20 checks not measured — two prompts got no answer from either engine, so they are in neither number above.

  • 25 AUG 2026 · 06:09 UTC

    Gained a citation on “top review management providers”

    ChatGPT cited example.com in 2 of 3 runs. The week before it named you without fetching anything, in 3 of 3.

  • 18 AUG 2026 · 06:11 UTC

    Partly measured — 6 of 20 checks did not run

    Three prompts came back from neither engine — that is the 6 checks, and it is why the week's numbers are computed over the 7 prompts that did answer rather than all ten. The denominator says so wherever it appears.

    reason: billing_exhausted — the ChatGPT account had no balance for part of the run.

  • 11 AUG 2026 · 06:10 UTC

    Nothing was measured

    Every check failed to return, so this scan produced no finding at all — not a low one, not a zero. It is left in the record because deleting it would make the line above join 4 August straight to 18 August and imply we were watching all along.

    reason: billing_exhausted — ChatGPT: the account balance was exhausted, 0 of 30 runs returned reason: upstream_error — Gemini: the provider returned 503 on every attempt, 0 of 30 runs returned
  • 4 AUG 2026 · 06:08 UTC

    Lost the mention on “how much does review software cost?”

    Both engines answered every time and named you in 0 of 3 runs each. That is a measured result: the answers came back and you were not in them. competitor-two.com was named in 3 of 3.

  • 28 JUL 2026 · 06:07 UTC

    First scan recorded

    Cited in 1 of 8 measured prompts, mentioned in 3 of 8. Everything above is measured against this.

Trend

6 scans · 28 Jul – 1 Sep 2026

Two lines, never one. A citation means your page was fetched and linked; a mention means a model named you from memory. Adding them together would claim the stronger evidence for the weaker.

Prompts where Northlight was cited, and prompts where it was mentioned, across six weekly scans Cited rises from 1 to 3 prompts and mentioned from 3 to 4 between 28 July and 1 September 2026. The scan of 11 August measured nothing, so both lines break there and the gap is marked "not measured" rather than plotted as zero. The number of prompts that could be measured varies between 7 and 9 and is printed under each date. 10 8 6 4 2 0 NOT MEASURED 28 JULof 8 4 AUGof 9 11 AUG— 18 AUGof 7 25 AUGof 9 1 SEPof 8
Cited — page fetched and linked Mentioned only — named from memory Not measured — the line stops

The y axis counts prompts, not percent. The denominator under each date is the number of prompts where at least one engine answered that week, and it moves — 7, 8 and 9 all appear above. A percentage would have hidden that. 11 August has no point on either line and no denominator, because nothing returned; it is a hole in the record, drawn as one.

Scan history

Every scan, including the ones that measured nothing.

Northlight · example.com · Manchester · 10 prompts × 2 engines × 3 runs
Scan Cited in Mentioned in Missing in Not measured Model
1 Sep 2026weekly · open 3 of 8 4 of 8 1 of 8 4 of 20 checks gpt-4.1-2025-04-14
gemini-2.5-flash
25 Aug 2026weekly 3 of 9 4 of 9 2 of 9 2 of 20 checks gpt-4.1-2025-04-14
gemini-2.5-flash
18 Aug 2026weekly 2 of 7 4 of 7 1 of 7 6 of 20 checks gpt-4.1-2025-04-14
gemini-2.5-flash
11 Aug 2026weekly not measured— not measured— not measured— 20 of 20 checks not reported
no answer came back to stamp
4 Aug 2026weekly 2 of 9 3 of 9 4 of 9 2 of 20 checks gpt-4.1-2025-04-14
gemini-2.5-flash
28 Jul 2026weekly · first 1 of 8 3 of 8 4 of 8 4 of 20 checks gpt-4.1-2025-04-14
gemini-2.5-flash

The denominator in the first three columns is the number of prompts where at least one engine answered, and it differs row to row. It is printed on every cell rather than in the header for that reason.

This brand

domainexample.com
location ManchesterA brand row is the brand at one location. Two outlets sharing a domain get two rows and two histories — collapsing them by name or by domain is how forty branches become one customer.
cadenceWeekly · next Mon 8 Sep
enginesChatGPT + GeminiThe pair whose per-call cost has actually been measured.
runs per prompt 3Every rate on this brand is k of n over those runs. AI answers vary run to run, so one run is not a number.

Coverage, last 6 scans

Of 120 prompt × engine checks scheduled since 28 July, 82 returned an answer.

38 checks
Did not run
counted here so the gap is visible, and excluded from every rate on this page

This is a count of what we could not see. It is not a score, and it is not subtracted from anything.

Why the log is above the chart

A number on a dashboard is read as current whether or not it is. A log entry carries the moment it was true and cannot go stale, so it goes first and the tiles carry an “as at” date. Stale numbers are worse than no numbers.