Each card in the feed now carries a number, like +4 pts, and the feed is sorted by it. The number is how much the AI visibility score would move once the change on that card is made. A list of observations becomes an order of work. Open the dashboard and the first card is the one worth the most, whatever it is about.
The number comes from a measurement. Every question in an analysis goes to ChatGPT, Google AI, Perplexity, Claude and Grok at once. When one assistant leaves a business out, the other four answers to the same question show what it could have said. Reachd takes the typical one of those answers, recomputes the score with the gap closed, and prints the difference. A card that names a page and a sentence is priced against what the same market already delivers on that question.
How the AI visibility score gain is measured
The AI visibility score is an average. Every answer in the scoring window gets a position score, first place highest, absent zero, and the score is the mean over all of them. That makes the value of a change a plain calculation. Take the answers a card is about and replace the empty ones with what the same question earned elsewhere. Average again, and the difference is the number on the badge.
Two kinds of gap get priced this way. The first is an assistant that misses what its peers recommend, say Claude leaving out a clinic that ChatGPT, Google AI and Perplexity all name. The second is a question the business keeps losing everywhere, on a topic the panel will ask again next week. Both are read from the same answers. Where no assistant has found the business on that question yet, the second kind falls back to the weakest ranked mention. A business that starts being recommended enters the ranking somewhere, and its bottom is the entry nothing can undercut.
Hover the badge and the card spells the sentence out: “Closing this adds about 4 points to your AI visibility score in Miami.” The market is named because the number is priced against that market’s own answers. Against the all-locations score the same change is worth proportionally less, since the denominator there is every market at once. The card says which one it means. A free AI visibility checker run shows the score this number is measured in.
Why the gain is priced on the typical AI recommendation, never the best one
The first version of the pricing took the best answer any assistant gave and assumed the missing one could match it. The cards read well and promised too much. We measured it on four businesses with the reading as the only variable, and the best-answer figure ran 2.4 to 6 times the typical-answer figure on every one of them. A multiple that consistent is one claim too large everywhere rather than noise. One card promised 16 points of a 29-point score for an assistant that had recommended the business in none of its 42 answers.
So the reachable value per answer is now the middle of what the other assistants delivered on that question. Four assistants agreeing on a business show what the fifth could say about it. One assistant on one lucky day shows much less.
The value of a change is what the same answers already achieve without this one gap. Neither a ceiling nor a floor, which is why the card says about and never at least.
AI visibility score points from different insights never add up
Two cards can describe overlapping answers. A gap on Claude in Cyprus and the same gap in Poland are two cards with two numbers. Closing both moves the all-locations score by less than their sum, because each was priced over its own market. So the product shows no total anywhere. Each number is what its own card is worth, read on its own market, and that is what lets two cards showing the same number mean the same thing.
How the feed ranks insights for AI search visibility
Priced cards come first, in order of points, discounted by how sure the pattern is. A card that can say what closing it is worth outranks one asking for work on a promise, whatever the second one’s urgency label says. Inside the priced tier, a point on ChatGPT or Google AI counts in full for the ordering. A point elsewhere counts a fraction of it. That weighting reflects where buyers ask, and it decides only which card leads. The score itself counts every assistant equally, and the score is what the card claims.
A card also has to be worth at least a point and a half on that weighted number to exist at all. Below it the gap is real but too small to spend a slot on. The feed keeps its six places for changes that would show on the chart. Cards the score cannot price yet, a directory where rivals are listed and the business is not, still appear and rank after the priced ones. The Italian Cabinet case shows what working down such a list looks like. Absent from every answer on March 11, the business was the most recommended of its Miami competitors six weeks later.
How this compares to other AI visibility tools
Recommendations with a label are common in the category. Otterly lets you filter its suggestions by Impact as Low, Medium or High. AthenaHQ shows a three-segment colored bar for impact against effort beside each insight. Writesonic’s Action Center scores Impact, Confidence and Effort, each as a level, and its full version sits on Enterprise. Profound keeps its Aim prioritization off the self-serve plans altogether, listed as None on both in August 2026.
A label answers whether a recommendation matters. A number answers by how much, in the unit the owner already watches. Two recommendations about different assistants and different pages become comparable on one scale. Reachd is the only tool in the category that prices a recommendation in points of the AI visibility score. The number is included on every plan, and it is on the first free analysis too, which the previous entry describes.
What it changes for getting recommended by ChatGPT and Google AI
The feed used to answer what is wrong. It now also answers what to do first, and by how much it matters. An owner with an hour a week opens the dashboard, takes the top card and makes the change on the page it names. The full platform already ranked its cards on this arithmetic underneath; the number on the card puts it in view.
The next runs report whether the points arrived, and once they do the card closes itself on that data, dated, with no box to tick.
Frequently Asked Questions
What does the number on an AI visibility insight card mean?
It is how many points the AI visibility score would gain if the change on that card were made. The score is an average over every answer in the scoring window, so Reachd recomputes that average with the missing answers filled in and prints the difference. The badge reads +4 pts, and hovering it spells the sentence out.
Does the number tell me how to get ChatGPT to recommend my business?
The card does, and the number says how much it matters. Each card names the assistant that leaves the business out, the question it loses, the page to work on and the wording that page is missing. The points beside it are what that change is worth on the AI visibility score, so the card at the top of the feed is the first thing to do.
How is the AI visibility score gain measured rather than estimated?
Every question goes to ChatGPT, Google AI, Perplexity, Claude and Grok in the same run. When one assistant leaves the business out, the other answers to the same question show what a recommendation there looks like. Reachd takes the typical one of them, never the best, and reprices the score with that answer in place.
Why does the card say about 4 points on the AI visibility score and never at least 4?
Because the value is read from what the same market already delivers without this one gap. It is neither a ceiling nor a floor. Answers to neighbouring questions would probably improve too, but nothing in the data says so, so they are left out of the number.
Can I add up the points of several insights?
No, and the product never shows a sum. Two cards can describe overlapping answers, so closing both moves the score by less than their total. Each number is what its own card is worth on its own market.
Why do some AI search insights have no number?
A card is priced only when the AI answers can show what it is worth. Placement cards, such as a directory where rivals are listed and the business is not, describe a gap the score cannot yet measure, so they carry no number and rank after the priced ones.
Do other AI visibility tools show the value of a recommendation?
Most show a label. Otterly filters recommendations by Impact as Low, Medium or High, AthenaHQ shows a three segment bar for impact against effort, and Writesonic's Action Center scores Impact, Confidence and Effort as levels. None prints the number of points a recommendation is worth on the score the owner watches.