The AI Citability Snapshot
Across the client websites we monitor, AI systems read content 140 times for every visitor they send. In July: more than 70,000 AI reads, fewer than 500 human visits from AI answers.
The numbers
What our research tells us about how the AI operators are reading your site:
Reads vs referrals
140
:1
Bot requests
3
in 5
Three of every five bot requests on those websites now come from AI systems, not search engines or SEO tools.
13 Sites
3
m+ hits
The monitoring base behind this service, with every ChatGPT, Anthropic and Perplexity hit checked against the operators' published server lists.
Figures from our client monitoring portfolio, July 2026, anonymised and aggregated. We publish the pattern, not the clients. This is what we call citability: whether the machines actually read your content and reach for it to answer real questions.
What you get for £1,000
Our monthly client AI reporting, applied to your website for one month.
The same data pipeline, the same verification rules, the same report format that we deliver to clients every month.
The written snapshot report.
01 - Report
Who is crawling you, which AI operators actually read your content, what they read, and what we would do about it. Typically 8 to 12 pages.
Access to the live dashboard.
02 - Dashboard
Your month in our Data Studio reporting, yours to explore for the duration of the engagement, so you can see the detail behind every headline number.
Verified numbers, not guesses.
03 - Validated data
Every ChatGPT and Perplexity hit checked against the operators' published IP lists. Anything we cannot verify is excluded, and we tell you what percentage that was.
A walkthrough call
04 - Feedback
45 minutes on the findings and the three moves we would make first.
Who is actually crawling you
The full bot split: AI training crawlers, AI search indexers, live assistants, search engines, SEO tools, and the bulk crawlers that are just noise.
Which AI operators your content reaches
Your reads by operator, and whether your machine readership rides on one operator or several.
AI reads versus visits
How often AI systems read your content against how many people arrive from an AI answer, with the honest caveat on what analytics can and cannot see.
What the machines actually read
Which sections and pages draw AI reading, and whether that matches what you want to be known for. On every website we monitor, it is the insight content, not the service pages.
Where you sit on the AI audience journey
Each operator placed on the five stages: train, index, read, represent, refer. Including the stage nobody can observe, shown as a gap rather than left out.
What we count and what we leave out
The verification tiers, what percentage of traffic we excluded as unverifiable, and what server logs cannot capture.
What we would do about it
Three moves, in order, written to be actionable whether or not you ever speak to us again.
How it works
One hour of your team's time. The rest is on us.
If your server already keeps 30 days of logs, we work from the month you have and deliver in about two weeks. If it does not, we switch retention on and report a month later.
01
A short setup call.
We agree how we get the logs: SFTP access, a log export, or your CDN's logs. Most platforms take under an hour of your team's time.
02
We ingest and verify.
Your month goes into the same BigQuery pipeline as our client monitoring. Every AI hit is identified, and ChatGPT and Perplexity traffic is checked against the operators' published server lists.
03
We write the report.
You get the written snapshot and access to the dashboard.
04
We walk you through it.
45 minutes on what it means and the three moves we would make.
What happens next?
Most organisations ask the same question next: what changed?
The snapshot continues as monthly reporting. Same report, same dashboard, delivered as a series so the trends become visible.
And when the question becomes "what should we change?", that is where our core engagement starts: mapping the audience you most need to win, human and machine, and building the content programme that serves it.
If you would rather start with an outside view than your own logs, our free Second Opinion assesses how one audience of your choice meets your organisation.
limitations
What a snapshot can't tell you.
A month is a snapshot, and the numbers move. A snapshot tells you where you stand; only the month-on-month series shows the trend.
Logs show reading, not representation. We can show you which AI systems read your content and how often a real question pulled a page in live. What the answers actually said about you happens inside the conversation, where no log reaches.
And the snapshot tells you what the machine audience is doing, not what to do for the audiences you care most about. That is audience work, and it is a different engagement.
FAQ
The practical bits
Our site sits behind Cloudflare or another CDN.
Then origin logs understate the true picture, sometimes substantially, and we would rather use the CDN's own logs. We will tell you which we are reading from and what it means for the numbers.
We don't keep server logs.
Most hosts can switch retention on in minutes. We help set it up on the setup call, and the snapshot covers the following month.
Is there anything sensitive in what you take?
We work on bot and crawler traffic. The pipeline is built for exactly this, and we agree scope and handling on the setup call.
Does this cover Google's AI Overviews?
Not from logs, because Google's AI answers run on the ordinary search index. If you grant us read access to Search Console, we include Google's own count of how often your pages appeared in AI Overviews and AI Mode, which pairs the appearance count with what the logs show being read.
One month of your server logs, decoded.
Tell us the website and anything you already know about your setup: host, CDN, whether logs are kept. We will come back within two working days with the setup call invitation.
£1,000, fixed.