Hook: Once a quarter someone senior asks whether the AI work is working, and a screenshot of a chat window from last Tuesday will not survive the meeting. Plot twelve months of AI visibility and export every page the models cited, work that costs EUR 3,240 to 6,480 a year to rebuild by hand.
What You'll Learn
- Plot a line with a date range on it, using the Rankfor Index or any single part of it, and drop a rival onto the same axes
- Read the seven parts the score is built from, each carrying the weight it contributed: Recommendation Share, Sentiment, Awareness, Consistency, Source Quality, Stability and Coverage
- Break sentiment out theme by theme, with the positive share of mentions and a signed score and mention count on each theme
- Open every graded citation split positive, neutral and negative, then filter down to the negative ones
- Export one image for the slide and one file for the sceptic, listing every cited page under its sentiment
Total Time: 5:33 | Difficulty: Beginner
The follow-up is the one that kills the meeting: fine, but what are they saying about us? A screenshot answers neither question. It does not line up against last month, and it names no source.
What it costs today
Assumption: an in-house analyst at EUR 90 an hour, rebuilding the monthly AI-visibility tracker by hand.
| Line item | Time per month | Cost per month |
|---|---|---|
| Re-run the prompt battery on the model | 1:00 to 2:00 | EUR 90 to 180 |
| Re-tally the answers by brand and position | 1:00 to 2:00 | EUR 90 to 180 |
| Patch the spreadsheet and redraw the chart | 1:00 to 2:00 | EUR 90 to 180 |
| Every month | 3:00 to 6:00 | EUR 270 to 540 |
| Every year | 36 to 72 hours | EUR 3,240 to 6,480 |
Every line in that table is asking and tallying, and none of it is scoring. Scoring each answer by hand, grouping the scores by theme, then grading every cited page is work almost nobody does, so what reaches the board is a mention count with the word sentiment on it. The method also drifts each time a different analyst takes over, which puts two different measurements on one axis. Marketer is EUR 196 a month, EUR 2,352 a year, and the method holds from month one to month twelve.
What the video shows
- The Rankfor Index on the project overview. One score, its band underneath, and a pill carrying the change since the last scan. The pill stays hidden until you have two.
- The seven parts under it. Hover any label and the tooltip says what it measures. Sentiment grades how positive the models sound about you, where fifty is neutral and a hundred is uniformly warm.
- Your Visibility through time. Pick the line from the dropdown, plot Sentiment on its own axis, set the window to a year.
- Put a rival on the same axes. Compare with drops a second project into the chart as a grey line. Anything never scanned on this model greys out and says why.
- The full Index view. Every tile shows the weight it contributed, so the total is arithmetic you can check, and a part the scan could not measure says so on its own tile.
- What they say, theme by theme. The Sentiment card gives the gauge, the positive share of mentions, and one row per theme with a signed value and a mention count. Bars run right for kind and left for harsh.
- Ten months on one screen. Historical data puts one bar per scan on the left and every component on one axis on the right.
- Which sources are saying it. Source and Language splits every graded citation into positive, neutral and negative, then filters to the negative ones. Only graded citations are counted, so those three numbers can come to less than the total above them. Open a row and each page carries its theme, its sentiment and its language.
- The file your CFO can audit. Export as Markdown for the Sources by Sentiment section, and as CSV for the raw sources table.
What you need
- Marketer, EUR 196 a month.
- Two or more completed scans on the same model. One scan draws no trend and no line.
- Strategist, EUR 496 a month, for the rival comparison, with the rival scanned on the same model. Free and Marketer hold one project per workspace; Strategist holds three.
Questions people ask
Why does the chart fill up so slowly? One point is one scan, and Rankfor.AI allows one scan per model per calendar month. Ten points is ten months. The cap is what keeps the points comparable.
Does Rankfor.AI find the themes in my own data? No. The themes are a fixed set carried on every prompt. Each answer is scored on its own, then grouped by theme, and each theme gets one vote, so twenty answers on one theme cannot outvote three on another. We publish no confidence levels and no margins of error.
Why don’t the screen and the file always agree? The chips filter the screen and every export carries the full citation list, so clear the chips before you export. A theme value shows one decimal on screen and two in the file, and a theme the card writes as governance reads Trust & Governance on a citation chip.
Will it warn me when the score falls? No. Rankfor.AI has no monitoring and no alerting, and there is no scheduled scan. You run the scan when you want the next point on the chart.
Set the window to a year, name the metric, and take the screenshot into the board pack. Start with Rankfor.AI.