What AI says about our own product
Before we run this for anyone else, we publish the same test on ourselves. This is the unedited result: our product, Is This Page Good, across nine AI engines: which recognise it, which describe it wrongly, and which have never heard of it.
Method
18 prompts per engine: 15 unbranded (a buyer asking for a page-review or landing-page-audit tool, never naming us) and 3 branded (asking directly about Is This Page Good). Cold sessions, no account history, no search pre-seeding. Each answer recorded verbatim and scored for (a) whether we were named unbranded, (b) whether the description was correct.
The result
| Engine | Named unbranded | Branded recall | Described correctly? |
|---|---|---|---|
| Inkling (Thinking Machines) | 2 of 15 | 3 of 3 | No — called it “professional critiques, not automated scores” |
| Inkling-small | 0 | 3 of 3 | No — misread the mechanism as “peer feedback” |
| Gemini 3.6 Flash | 0 | Named | Yes — accurate description |
| DeepSeek Chat | 0 | 0 of 3 | Never heard of it |
| GLM 5.3 Flash | 0 | Name only | “No knowledge” |
| Nex 2.5 Pro | 0 | Name only | “Can't verify” |
| Laguna S 2.1 | ~11 of 15 (flagged) | — | Anomaly — pending re-probe |
Baseline run 2026-09-14. Rows marked “flagged”/“anomaly” are excluded from the headline rate until re-probed.
Where it stands now — three weeks in
We re-ran the same eighteen questions on 2026-10-07, about three weeks after the run above — same questions, plus ChatGPT, which wasn't in the first run.
| Engine | Knows the name | Describes it correctly? | Named unprompted |
|---|---|---|---|
| ChatGPT | 3 of 3 | Yes | 0 of 15 |
| Perplexity | 3 of 3 | Yes | 0 of 15 |
| Claude | 3 of 3 | Yes | 0 of 15 |
| Gemini | 3 of 3 | Yes | 0 of 15 |
Re-run 2026-10-07. GLM and Laguna repeated the name but stated no knowledge of it; Gemma was unavailable. The original panel's smaller experimental models are partly retired, so this is the same eighteen questions on the engines that still run — plus ChatGPT.
What moved: three weeks ago, of the engines that knew the name, most described it wrong — a peer-review community, “professional critiques, not automated scores.” Today the four engines that matter describe it accurately: an AI landing-page audit tool from Memeador, a free report, about ninety seconds.
What hasn't changed: when nobody asks our name, we're still not in the answer. That part is earned off-site, over months — which is exactly what the sprint is for.
What it says
- Unbranded citation rate: about 2%. One engine, two prompts, and always as a runner-up next to a competitor.
- Branded recall is uneven: four of seven engines recognise the name, and three of those describe it wrongly. A wrong description spreads as easily as silence.
- The engines name our competitors unprompted: Hotjar, Microsoft Clarity, GA4, Unbounce, CXL, Crazy Egg, VWO, Maze, Attention Insight.
- The gap is quotable facts. The engines never say what our product is because our pages never said it in a sentence a machine can lift. That is the fix — and it's exactly what we sell.
We'll run the same test on you.
Three real buyer questions, the answers verbatim, and the competitor named instead of you. Free.
Get the free mirror test →