Measure
August 6, 2026 · 4 min read
By Marcus Bransbury · Founder, Robot Visible
Why Perplexity did not cite our website
A dated Perplexity check: the exact question, verbatim answer, 20 sources, and seven tools it named instead of us—plus what that result can support.
Quick answers
Why publish a check that you lost?
Because a case study that only exists when the result is good is marketing, not measurement. The product's central claim is that you should record flat and declined outcomes the same way you record wins, and the first place to demonstrate that is on ourselves. It is also the most honest way to show what a first measurement usually looks like.
Can I reproduce this?
Yes, and that is the point of publishing the question, provider, model, surface and timestamp. Ask Perplexity the same question through the Sonar API and you will get an answer of your own. It may well differ from ours: answers vary by time, surface, location and personalisation, which is exactly why one result is a sample rather than a trend.
Does being absent from one answer mean the site has a problem?
It means one specific thing: for that question, on that provider, at that moment, the retrieval step did not surface any of your pages. That is worth acting on when the question matters to your buyers. It is not a verdict on the site as a whole, and it is not the same as a readiness failure — ours passed every readiness check while producing this result.
The check, in full
On 6 August 2026 we asked Perplexity, through the Sonar API on the `sonar` model, a question a real buyer would ask: “What tools can check whether AI assistants like ChatGPT and Perplexity recommend my website?” That is our category question. If there is one answer we should be in, it is that one.1
Robot Visible was not mentioned, not recommended, and not cited. The answer named seven other tools. It drew on twenty sources, and not one of them was robotvisible.com.
| Field | Value |
|---|---|
| Question | What tools can check whether AI assistants like ChatGPT and Perplexity recommend my website? |
| Provider and surface | Perplexity, Sonar API, model sonar |
| Asked at | 6 August 2026, 08:43 UTC |
| Our result | Not mentioned, not recommended, not cited |
| Sources the answer used | 20, none of them ours |
| Other tools named | 7 |
Who was in the answer instead
The answer put forward Citation Rocket, Search Engine Land's AI Agent Readiness Checker, Found by AI, friendly4AI's AI Visibility Score, AI Exposure Tool, CiteRank, and UltraScout. Our own extraction ranked three of them highest: CiteRank with the strongest evidence, because the answer cited citerank.io directly, and Found by AI and AI Exposure Tool at lower confidence, because the answer recommended them without citing their sites.
That list is what one answer said on one date. It is not our assessment of those products, and being named by an answer engine is not a judgement about quality — it is a measurement of retrieval and phrasing.
Readiness was not the problem
This is the part worth sitting with. Our site is fetchable, indexable, permitted for every named AI crawler, and carries structured data. It passes the checks our own scanner runs. None of that put us in the answer, because none of it is the same measurement.
That gap is the entire reason this product exists, and running the check on ourselves is the cheapest possible demonstration of it.
A readable site can still be absent from every answer›
What this result licenses us to say
- That on this date, on this surface, for this question, Perplexity did not name us. That is a fact.
- That the answer used twenty pages and none of ours was among them, which points at retrieval rather than phrasing as the first thing to fix.
- That seven competitors were named, and one of them had its own site cited.
And what it does not
- It is not a trend. One answer at one moment is a sample, and answers vary by provider, surface, time, location and personalisation.
- It says nothing about ChatGPT, Claude or Gemini, because we did not ask them. A provider we did not check is not a negative result.
- It is not a ranking of the tools it named, and it is not evidence that they are better than us at anything except appearing in this answer.
- There is no “after” yet. We have not shipped a change against it, so there is nothing to compare. When there is, the comparison will state the interval and the result, including if it is flat.
The next move it points at
The answer was assembled from pages that name specific tools and say what each one does. Our own comparison content exists but was not retrieved for this phrasing. So the change this evidence selects is not “write more content” — it is to make the pages we already have retrievable for the question as it was actually asked.
That is one change, chosen by one observation. We will re-run the same question, on the same surface, after it is live, and publish what came back.
Sources and further reading
- Sonar API — Perplexity. Documents the API surface and models used for this check, which is what makes the provider and surface reproducible rather than a claim about the consumer product.
- A new resource for optimizing for AI features — Google Search Central. Supports the finding that established search fundamentals govern retrieval, which is where this check located the problem rather than in the phrasing of the page.
Continue learning
See where your website stands
Run a free scan and get your AI readiness score across all six categories, with the gaps to fix first.