Monitoring
Find out who the AI recommends when someone asks for a firm like yours.
Your buyers are asking ChatGPT, Gemini and Perplexity for a shortlist before they ever reach your website. Someone gets named in that answer. We tell you whether it is you, who it is instead, and for which questions.
Today
"Which firm should I use for cross-border employment contracts?"
Three providers are commonly recommended for this: a competitor, a competitor, and a directory listing.
You are not in the answer.
The goal
"Which firm should I use for cross-border employment contracts?"
Providers frequently cited on this topic include your firm, alongside two others.
Named, with a source we can point to.
Six engines, two methods
Not every engine can be measured the same way.
Half the engines search the web when they answer. Half answer from what the model already knows. Those produce different evidence, so we report them differently instead of averaging them into one flattering number.
01
Live search
Perplexity, Gemini and Google AI Overviews retrieve pages at answer time. We capture the source URLs they actually returned, so a citation can be traced to a specific page.
Evidence: source URLs and domains
02
Training data
ChatGPT, Claude and Copilot answer from model knowledge, with no live retrieval. There are no URLs to capture, so we record whether you were named and treat that as a weaker signal — because it is one.
Evidence: named or not named
Attribution
A mention is not a citation until we can prove where it came from.
When an engine names you, we try to resolve that back to something we published, in strict order. The method that resolved it is recorded on every row, so you can always see how strong the evidence is.
→
The ladder
An attribution token embedded in the published URL is the strongest signal. Failing that, an exact URL match, then a scored content fingerprint. Anything resolved only by query or ambient brand mention is recorded but never counted as an earned citation.
Recorded on every monitored answer
Why measure every engine
Being visible on one AI surface tells you very little about the others.
Source overlap between engines
<0.2
Average Jaccard similarity between the sources retrieved by Google Search, Gemini and AI Overviews — they barely agree on who to cite.
Grossman et al., ACM SIGIR 2026
Citations that support their sentence
74.5%
Only three-quarters of citations in generative search engines actually supported the statement attached to them, which is why we verify rather than trust.
Liu, Zhang & Liang, Findings of EMNLP 2023
Checks per month on Growth
60
Monitored questions per month across all six engines, with the full per-query source log kept for every run.
Growth plan · €199/month annual
Questions people actually ask
Differently, and we label which is which. Perplexity, Gemini and Google AI Overviews answer from a live web search, so we capture the source URLs they actually returned. ChatGPT, Claude and Copilot answer from model knowledge, so there are no URLs to capture and we record whether your business was named. Treating those two as the same number would flatter the report and mislead you.
An earned one. A mention only counts when it is tied to a specific article of yours and resolved by URL, by attribution token, or by content fingerprint. Brand mentions we cannot trace back to something we published are recorded in the query log, but they never count toward your citation number or trigger a first-citation alert.
No, and nobody honestly can. Model responses are stochastic: the same question can return different sources on consecutive runs. What we can do is run the work that raises the probability, measure what actually happens across six engines, and show you the evidence either way.
Indexing is fast and citation is slow. Public content is typically reindexed within days to weeks, but a stable, repeatable presence in AI answers usually takes months of continuous work. Any vendor quoting you a precise week count is guessing.
Yes. Every monitored query records which domains were cited, not just whether you were. That is how you find the questions where someone else is being recommended and you are not.
See which answers name you today.
A free analysis of your domain across the engines your buyers use. No card, no call required.
✓ Free✓ No card✓ Two minutes✓ GDPR
