How we measure your visibility in AI answers

We ask your customers' questions in their own wording: a full sentence in your market's language and dialect, not a keyword. The question is the unit here, because an engine answers questions and does not rank words.

Every site carries its own country and dialect, set before the first run. Both change the answer: a customer in Riyadh and a customer in Dubai ask in different words, and Google's AI answers differ from one country to the next. So your questions are written for your market, Google's answers are read as they appear in your site's country, and what you get is your market rather than a generic version.

Questions come from three places, and you can mix them. Start with buying questions from a customer close to deciding, not general questions about your industry. Choosing the questions worth measuring works through it with examples.

• You write them: the questions you hear in sales and support every week.
• We suggest them from Search Console: real searches that already reach your site, if you connect it.
• We suggest them from your pages: when Search Console is not connected, or you start from nothing.

We ask 8 engines: ChatGPT, Google AI Overviews, Gemini, Perplexity, Google AI Mode, Copilot, Claude and Grok. How many a measurement uses follows your plan, and the free plan starts with 2.

They fall into two kinds, and the difference is how your customer reaches them:

That difference is not a detail. A chat engine answers one person in one moment; a search surface shows the same page to everyone who searched that hour. What sets each engine apart is in the engines we measure, and which your plan includes is in Tasadr plans and prices. How each engine picks its sources is in the ChatGPT guide in Arabic and the guide to Google AI Overviews in Saudi Arabia.

• Chat engines, being ChatGPT, Gemini, Perplexity, Claude and Grok: your customer opens them and asks, and the answer moves between one ask and the next.
• Search surfaces, being Google AI Overviews, Google AI Mode and Copilot: they reach your customer mid-search, inside the results page itself, in Arabic and in their country.

We measure every question again on every run, so one lucky answer is never counted as a change. That is the difference between knowing what an engine said once and knowing what it usually says.

How often a new analysis arrives follows your plan, and it runs without anyone pressing anything:

A denser picture is bought with cadence: the daily plan gives you a fresh reading every day, so you read the week as a line rather than a point. What each plan carries is set out in full on Tasadr plans and prices.

• Free: a new analysis every month.
• Starter and Pro: a new analysis every week.
• Business: a new analysis every day.
• On-demand runs: start one yourself whenever you want, and your plan sets how many: 1 on-demand run a month on Free, 4 on-demand runs on Starter, 10 on-demand runs on Pro and 30 on-demand runs on Business.

For every question on every engine we record one of four outcomes, and keep the answer text as written with the links it cited, so you can check it rather than believe us:

The difference between absent and no answer is not a detail. The first means the engine chose someone else; the second means it chose nobody. That is why no answer is never counted against your site: confusing the two makes the score lie in both directions.

• Cited: your site named with its link inside the answer, the best outcome there is.
• Mentioned: your brand in the text, with no link to open.
• Absent: an answer was written, naming someone else or nobody.
• No answer: the engine wrote nothing for that question on that run.

The score is one number out of 100 saying where your site stands inside the engines' answers to your questions, per site. You read it in a line, then read under it the question you won and the question you lost, because a single number hides both.

When a measurement spans a range, we show it beside the score: the lowest and highest figure in that run. A wide range means the number is less settled than it looks, and your decision needs that.

An engine that wrote no answer is never counted against your site. It chose nobody, and charging you for its silence would lower your number over an answer that was never written.

Under every verdict sits the evidence: the answer text as the engine wrote it, the links it cited, and the date. The score is not an opinion you take on trust, it is the summary of texts you can open and read yourself.

When the change is confirmed, not on a single answer that carries it. The rule is written in advance and is the same for every account, so headlines and alerts do not flicker:

What passes the rule reaches your dashboard as an alert; what does not stays in the history for you to read. So you never wake to an alert caused by one odd answer.

• A new rival: we do not call it new on a single answer; the change has to hold.
• A lost citation: we do not report it lost on a single answer without it; the absence has to hold.
• A new gain: counts at once, because a citation that happened has happened.

For every question we record who was named instead of you and how often, then compare your page with theirs: what their page answers that yours does not. How many competitor analyses you get a month follows your plan: 3 competitor analyses on Free, 15 competitor analyses on Starter, 50 competitor analyses on Pro, and unlimited on Business.

The repeated gaps become an action plan ordered by priority: what to change first, on which page, and why. Ordered by expected effect rather than frequency, because one large question can be worth ten small ones.

The comparison itself is plain: your page and theirs side by side, and what each answers. We do not guess at the engine's hidden rules: we show what is on their page and missing from yours, and you decide which gap is worth an hour. How to read that comparison and follow a rival over time is in which competitor ChatGPT recommends instead of you.

We built our method on the Arabic question from day one. It has seven points, and their value is in the combination rather than in any one of them:

1. 1Questions in your customer's words, their dialect and their own market, not the language of your internal deck.
2. 2Google's answers as your customer sees them in their country, not a generic version of another market.
3. 3Repeated measurement that separates luck from fact, so one lucky answer is never counted as a change.
4. 4Text evidence under every verdict: the answer as written, its sources, with its date.
5. 5An analysis that says why your rival was cited, and what to change on your page.
6. 6An alert only when a change is confirmed, never an alert for every wobble.
7. 7Clear prices in riyal, with no meter running on each question or request.

We say this early so you read the number for what it is: a high score does not guarantee visits, and a measurement is a picture of your questions, not a census of every conversation in the world.

What we do not sell yet, we also say: no PDF export, no team seats, no API, no agency-branded reports. For another question about the method or the plans, more than 20 answers sit in the questions people ask about Tasadr, and the field is defined in what GEO is.

• The score is ours: computed from our own measurements by a fixed rule, and issued by no engine.
• No citation is guaranteed: an engine picks its sources afresh every time, and nobody sells a place inside its answer.
• We do not measure what you did not ask: widening the question list is what widens the picture.
• We do not measure your traffic: we measure your place in the answer, and traffic is what Search Console is for.