Every client gets a baseline score and a branded report link. Six decks a month by hand costs EUR 5,760 to EUR 11,520.
Your client wants to know whether the retainer is working. That question sits behind every renewal call, and answering it is what they actually buy from you. AEO and GEO have no Search Console, no rank tracker everyone agrees on, no number a client already trusts. So you build the evidence yourself, per client, every month: write the prompts, run them, tally who got named, pull the logs, write the narrative, recolour the deck. Then the next client. The work that proves the work has grown bigger than the work.
What it costs today
In-house and analyst time at EUR 90 an hour, agency time at EUR 120. One client, one month.
| Job, by hand | Arithmetic | Cost |
|---|---|---|
| Does AI name them at all: 25 prompts, three models, run and tally | 7:00 first round, then 5 h a month, at EUR 90/h | EUR 630, then EUR 450 a month |
| Map Search Console queries to themes and fan them into buyer phrasings | 8 to 16 h at EUR 90 | EUR 720 to EUR 1,440 a month |
| Get AI crawler hits and AI-referred clicks out of the logs | 4 to 16 h to build, 2 to 4 h a month to keep, at EUR 90 | EUR 360 to EUR 1,440, then EUR 180 to EUR 360 a month |
| Follow one buyer question across three models | 2 to 4 h at EUR 90 | EUR 180 to EUR 360 per question |
| Question research, one market and one language | 8 to 24 h at EUR 90 | EUR 720 to EUR 2,160 |
| Pull the numbers, build the charts, write it up, brand the deck | 8 to 16 h at EUR 120 | EUR 960 to EUR 1,920 per cycle |
| The deck line across six clients a month | 6 × (EUR 960 to EUR 1,920) | EUR 5,760 to EUR 11,520 |
The report line alone runs against an invoice of EUR 1,500 to EUR 4,000. Six clients a month puts EUR 5,760 to EUR 11,520 of your team’s time into documents. The white-label reporting licence beside it, EUR 100 to EUR 400 a month, produces none of these metrics.
Marketer is EUR 196 a month. Discovery Lab and a roster of client projects in one workspace need Strategist at EUR 496.
The workflow
Six videos, 36 minutes, in the order you would run a new client.
1. Does AI name you at all? The baseline you open the engagement with. One score out of 100, four named metrics under it, and the prompt-by-prompt evidence behind Brand Recall: every question, the buyer type that asked it, the model’s answer. Then the list of brands the model recommended while your client was absent, with a count on each. Export as PDF for the board pack and CSV for whoever argues with the method.
2. Turn Search Console queries into AI prompts Link the client’s property and their search programme stops being a separate report. Every landing page gets matched to the closest territory on their own Brand DNA Map, with a similarity score per row, and no competitor in that comparison. Promote a real query to a persona and it becomes a prompt every later scan measures, badged Custom.
3. Which AI bots read you, and who clicks through One script tag before the closing body tag, and the client’s developer is done. After that: requests from AI crawlers, seventeen user agents collapsed into five company names, and clicks from people who arrived out of an AI conversation, split per chat, and the ten pages those answers send people to.
4. Follow one buyer question across three AI models One question to Gemini, ChatGPT and Grok in one run. You see the searches each model fired before it wrote a word, the pages it read, and a verdict per brand: Survived, Dropped, Emerged, Absent. Dropped is a proof problem, Absent is a coverage problem, and that distinction is most of your recommendation. It ends in a share link your client opens with no account.
5. Find the questions you never thought to track The prompt set your client approved is a list of questions their team already knew to ask. This run writes the ones nobody wrote down, asks them, and ranks every brand the model named back. The defaults produce 45 answers; the ceiling is 20 prompts, 5 rewrites each, asked 10 times, so 1,200.
6. Hand over a client report without building a deck The deliverable. Your logo, your accent colour, their team name, a six-digit PIN, one link. A PDF of the same page for the stakeholder who wants a file, and a sources CSV so “how was this number built” gets answered once. Branding saves against the project, so you set it once.
What this does not do
One project per workspace on Free and Marketer. A roster of clients needs Strategist at EUR 496 a month.
One scan per model per calendar month. The server refuses the second one. A take you would rather redo waits until next month, so build the reporting cycle around that date.
No monitoring and no alerting. Nothing watches a client’s number between scans and nothing emails you when it moves. You open the tab and read what came back.
Gemini, ChatGPT and Grok. Claude and Perplexity sit outside the scans and we publish no numbers for them. They do appear on the traffic screen as places people clicked from and as crawler names, which is what a script on a page can see.
Start here
Watch Follow one buyer question across three AI models first. It carries no tier gate, so you can run it tonight on a prospect’s brand before anyone has signed anything, and it finishes as a link that opens with no account. Two to four hours of analyst work per question at EUR 90 an hour is what you stop doing. A traced answer you can open in the meeting is what you start selling.
Then watch Hand over a client report without building a deck, because that link is what the invoice is for.