Hook: The board wants to know whether a buyer typing your category question into ChatGPT or Gemini gets your name back. One scan answers it with the prompts behind the score, and the same tally done by hand costs EUR 630 a round across three models.
What You'll Learn
- Register a brand and answer the seven onboarding questions once, then let the scan run on its own
- Read your AI Visibility score out of 100 with its band beside it, and the four metrics under it: Brand Recall, Message Alignment, Value Understanding, Content Structure
- Open the prompt-by-prompt evidence behind Brand Recall: every question, the buyer type that asked it, the model’s answer, and a Brand found tag where your name appeared
- See the brands the model named in answers to your buyers’ questions, with a mention count on each one
- Export the screen as PDF, Markdown or CSV so the board pack carries the method with it
Total Time: 5:43 | Difficulty: Beginner
Nobody in the building can answer that board question today, and the answer moves on its own between one meeting and the next. So somebody builds a spreadsheet: twenty-five prompts, twenty-five answers pasted in by hand, then a tally of who got named and who got named in your place.
What it costs today
Assumption: an in-house marketer loaded at EUR 90 an hour. The prompt set gets written once and reused; the run and the tally repeat for every model.
| Line item | Time | Cost |
|---|---|---|
| Write 25 buyer-intent prompts (once) | 2:00 | EUR 180 |
| Run 25 prompts on one model, log every answer at 2 min each | 0:50 | EUR 75 |
| Tally who got named, at what position, and which rival took the slot | 0:50 | EUR 75 |
| One model, first round | 3:40 | EUR 330 |
| Three models, first round (write once, run and tally three times) | 7:00 | EUR 630 |
One round shows a level and no direction, so the run and the tally repeat every month: 5 hours a month across three models, EUR 450 a month at EUR 90 an hour, EUR 5,400 a year. Agencies quote a one-off AI visibility audit at EUR 3,000 to 8,000 and hand back a PDF that ages the week your website changes. Marketer is EUR 196 a month, EUR 2,352 a year, and the run repeats itself.
What the video shows
- Register the brand. Name the project, paste the website address, tick Run scan for this project. The competitor box stays unticked, because ticking it skips the questions that tell the crawler who you sell to.
- Seven questions, once per project. Brand name and the ways people misspell it, industry, languages, markets, audience, the rivals you think you lose to, goals.
- The scan runs on its own. It reads up to a hundred pages of your website, works out the topics, writes the buyer types those topics imply, turns them into questions and asks the model. Four phases, and it keeps going when you switch tabs.
- The project opens. Everything on the overview hangs off that one scan. Go left into Personas, then AI Visibility.
- The score card. One number out of 100 and the band pill beside it, running Starter, Discovery Phase, Foundation Builder, Rising Contender, Market Leader. Under it sit the four metric cards in order: Brand Recall is whether you get named at all, Message Alignment is whether the model describes you the way you describe yourself, Value Understanding is whether it can say why you are worth buying, Content Structure is whether your pages can be cited.
- Prompt by prompt. Click the mention count and every prompt opens: the question and the buyer type on the left, the model’s answer on the right, scored, with a Brand found tag where your name appeared.
- Who got named in your place. Click Competitors mentioned and read the brands the model recommended while you were absent, each with a mention count.
- Take it with you. Export as PDF for the board pack and CSV for whoever wants to argue with the method.
What you need
- Marketer, EUR 196 a month.
- Your website address, and five minutes for the seven questions. You answer them once per project.
- One AI model picked for the scan. Rankfor.AI works with Gemini, ChatGPT and Grok.
- On Free the scan reads 10 pages, writes 1 buyer type and asks 1 question, so the Brand Recall list comes back as a single row. Paid scans read up to 100 pages, write 5 buyer types and ask 5 questions.
Questions people ask
How often can I re-run the scan? Once per model per calendar month. The server refuses a second one inside the same month, so three models means three scans a month and no more. Plan the month around it.
Which AI models does Rankfor.AI ask? Gemini, ChatGPT and Grok. Claude and Perplexity are outside the picker, and we publish no numbers for them.
Will it tell me the moment my score drops? No. Rankfor.AI has no monitoring and no alerting. You run the scan, you read what came back, and you run it again next month.
Do I get a trend line from the first scan? No. A trend needs two scans on the same model, so the chart reads “Not enough history yet” until the second one lands. The score, the prompts and the competitor list are all there from day one.
Run the seven questions on your own website address and read what the model gives back. Start with Rankfor.AI.