A sponsored comparison on a well-known site is bought on a promise: that ChatGPT, Perplexity, Gemini and Google's AI answers will cite it and name you. We measure that before the article goes live and again at 14, 30 and 60 days — repeated questions, control questions, an interval on every number, and a plain verdict.
Request a placement audit How it is measured
€249 per placement, one-off · we measure, we don't sell placements · no uplift promised · Version française
In two large published studies of AI answers to buying questions, “best X” list pages made up about 41–44% of the cited sources. The tactic is not imaginary — that is why it is being sold.
A public audit of Google AI Overviews found that when a self-published ranking was cited, the assistant usually recommended someone else. An article can be the source of an answer that names your competitor.
Ask the same assistant the same question twice and the list of brands often changes. “#1 in ChatGPT” from a single conversation says nothing about the next thousand.
A site the assistants already cite does not make a new page on it cited. It has to be crawled, indexed, retrieved and chosen — and it lives on someone else's domain, under their policies, for as long as they keep it.
| Measured placement audit | Typical placement report | |
|---|---|---|
| Before publication | A baseline on the questions the article targets, plus control questions it does not | Rarely provided |
| Repeats | Every question asked three times per assistant at every checkpoint | Once, if at all |
| What counts as success | You are recommended in the answer — reported separately from the article being cited | “The article is live” or “it ranks” |
| Drift | Control questions show what moved on its own | Mixed in with the article's effect |
| Result | Changed, within noise, or not picked up — with a 95% interval | A screenshot |
Up to 12 questions the article is meant to win, in your customers' words and language — written with you, or drafted from your site for your approval — plus 6 control questions from the same category that the article does not target.
Five assistants — ChatGPT, Perplexity, Gemini, Google AI Overviews and Google AI Mode — and every question asked three times: up to 270 answers. We keep every answer, every cited URL and every brand named.
Can the answer engines' crawlers reach the host site? Is the article indexed? How is it labelled, and what rel attribute do its links carry? Recorded at every checkpoint, because pages get moved, noindexed or taken down.
Same questions, same assistants, same repeats. Day 14 shows whether the article is picked up at all, day 30 whether it changes the answers, day 60 whether that holds.
For every answer: was the article cited as a source? Were you named or recommended? Was your own site cited? Who was recommended instead? Each outcome gets its own rate and interval.
The change on the target questions minus the change on the controls, with a 95% interval, per assistant. If the interval includes zero the report says “within noise”, in those words. The smallest change the sample can detect is stated with the baseline.
Before the article goes live: your current citation and recommendation rates, who is recommended instead, and which sources the assistants already use.
At 14, 30 and 60 days after publication, on the same questions, with the state of the article and the host at each date.
“Article cited in X% of answers; you recommended in Y%; change against control Z points” — each with its interval. Or “within noise”, or “not picked up”.
The brands the assistants recommend in your place, per assistant, and the pages they took it from.
A CSV row per question, assistant, repeat and cited URL. Nothing behind a figure is hidden from you.
A print and PDF layout; for agencies, under your name and colour.
Send it through the intake (choose “Placement audit”) or write to hello@answerrange.com.
Prices in euro, excluding VAT. Agencies auditing several placements: volume pricing on request. Order before the article goes live — a baseline cannot be measured after publication.
rel="sponsored" (or nofollow). Under EU consumer law an undisclosed paid advertorial is a banned practice, and the EU Court of Justice has held that the advertiser who paid for it can be responsible. Google's spam policies treat unqualified paid links as link spam and, since May 2026, also name attempts to manipulate AI answers in Search. The audit records the label and the link attributes at every checkpoint. This is not legal advice.A sponsored article is a media buy. Ask for the cited-before and the cited-after; if nobody can give you them, we will measure them.