Skip to contentSomeone above you is going to ask what you’re doing about ChatGPT. Walk in with a mention rate you can defend, a roadmap scored on demand, and the AI-referred traffic your analytics files under “Direct.”
WE ASK CHATGPT, PERPLEXITY, GEMINI AND AI OVERVIEWS WHAT YOUR BUYERS ASK, EVERY DAY, AND TURN THE ANSWERS INTO A NUMBER YOUR QUARTERLY REPORT CAN CARRY
The first time you report a flat week honestly is the last time anyone doubts your up weeks
Every marketing team gets this question this year, from the CMO, the founder, or the board. The common answer is a screenshot of one good ChatGPT run, and it survives until someone else runs the same prompt and gets a different answer.
Ask an engine the same question ten times and your brand can appear anywhere from 20% to 80% of the time. A screenshot is an anecdote. What upstairs wants is a number that behaves like the other numbers in your report: measured the same way every time, comparable month to month, honest about its own noise.
So that is what we built. A mention rate per prompt, sampled multi-run, every day, with a 95% confidence band, and a straight “no significant change” when nothing moved, so the changes you do report are believed.
Baseline. Lever. Attribution.
You already report organic, paid, email and social. AI answers are the channel forming underneath all of them: buyers ask an engine, get a shortlist, and never see the SERP your SEO tools measure. Three things earn a channel its budget.
A baseline
Where you stand today, per prompt, per engine, with the position you hold in the answer.
A lever
A weekly roadmap scored on demand against the visibility you have, grounded in the domains the engines cite for your topics, so the next content sprint has an ordered backlog rather than a hunch.
Attribution
The visitors AI sent you, currently buried in “Direct,” labelled certain or likely and never mixed. Plus verified reads by GPTBot, ClaudeBot and PerplexityBot on your pages, joined to what got cited.
Those three are what turns “we’re keeping an eye on AI” into a line in the quarterly report.
Where you stand
Mention rate across your selected engines, multi-run, daily, with confidence bands and the position you hold in the answer. The baseline slide, maintained for you.
Who is beating you
The rivals AI recommends instead, the prompts where they win and you are absent, and the domains those answers cite.
What to fix next
A weekly roadmap ranked by demand against current visibility. An ordered backlog for the next content sprint.
Whether it worked
Every prompt bucketed: winning, contested, losing, or too early to call. When movement clears the band it is real, and it is your slide.
Who AI sent you
AI-referred sessions your analytics files under “Direct,” labelled certain or likely. The traffic you are already earning without the credit.
What agents read
Verified crawler reads per URL. When GPTBot reads the new comparison page you will know it was read, and whether it was later cited.
If you run SEO
Your rank trackers cannot see the answer above the links. Track the prompts your keywords became, watch which of your pages the engines cite, and catch verified AI crawler reads per URL: the crawl report you already know how to read, pointed at the engines that matter now.
If you run content
The roadmap tells you what to write before you write it: prompts with demand where you are absent, and the domains the engines cite for that topic, which is your pitch list and your brief in one. Then the buckets tell you whether the piece moved anything, with a band around the answer.
If you run brand or PR
See which facets of you the engines praise and criticise, the integration, the pricing, the support, scored across your selected engines. That reading is taken from answers where you are named, deliberately, because those are the answers that describe you. Alongside it sits the set of domains those answers cite, which is where earned coverage would have to land to change the description.
Five minutes to set up. A briefing every Monday.
- .01
Tell us your brand
Setup >
Paste your URL. We draft the questions your buyers ask AI; you approve the ones worth tracking. Five minutes.
- .02
We interview the engines
Measure >
Every day, every tracked prompt, several runs each, because answers change and one run proves nothing.
- .03
Add the snippet
Detect >
One line of code for AI-referred visitors, plus a drop-in server-side helper so agent reads get counted, not guessed.
- .04
Read Monday’s briefing
Act >
What moved, who overtook you, which pages got cited, and three things to do about it. In a week with nothing measured we send nothing.
We lead with our receipts, not a hype score. So can you.
MULTI-RUN SAMPLING
Every tracked prompt, every selected engine, every day. We report the rate, never the lucky run.
95% CONFIDENCE BANDS
And “no significant change” when the bands overlap. Report noise as noise and your signal gets believed.
VERIFIED CRAWLERS
Every bot hit checked against published IP ranges. Impostors get their own bucket.
EU-FIRST, NO RAW IPS
EU hosting, no raw IP addresses stored, no cross-site tracking. Legal signs off without a meeting.
A marketing-team-fit AI visibility tool needs three things: a defensible metric, an actionable roadmap, and attribution. visitedby.ai measures brand mentions across ChatGPT, Perplexity, Gemini and Google AI Overviews with multi-run daily sampling and 95% confidence bands, delivers a weekly prioritised roadmap, and labels the AI-referred visitors that analytics tools file under “Direct.”
Not with one manual check. The same question asked ten times can name a brand anywhere from 20% to 80% of the time. Reliable measurement means sampling each prompt multiple times daily and reporting a rate with a confidence band, which is what visitedby.ai does across four engines.
Report a mention rate per tracked prompt with a confidence band, movement bucketed as winning, contested, losing, or too early to call, and AI-referred traffic as a labelled segment. Rates with bands behave like the other metrics in a quarterly report: comparable, repeatable, and honest about their own noise.
Yes. The weekly roadmap ranks fixes by demand against your current visibility and shows the domains the engines cite for each topic, so you know what to write and where a citation would have to come from.
Yes, and much of it lands in “Direct” in your analytics, because answer-engine referrals often arrive without referrer data. We label AI-referred sessions as certain or likely, never mixed, so you can size the channel honestly.
No. Your SEO tools measure the links; we measure the answer above them, which they cannot see. Your analytics counts sessions; we label the ones AI sent you and add verified AI-crawler reads per page.
Setup is five minutes and one snippet. After that it is a Monday briefing: what moved, who overtook you, and what to do next. In a week with nothing measured, we send nothing.
Built EU-first: EU hosting, no raw IPs stored, no cross-site tracking. The crawler-analytics side processes no personal data at all.
YOUR ENGINES, INTERVIEWED DAILY. A RATE WITH A BAND. A ROADMAP. THE TRAFFIC “DIRECT” WAS HIDING.
