What is the best tool to analyse your place in AI answers?

A GEO tool is software that asks AI engines the questions your customers ask, then records for every answer whether the engine cited your site with a link, mentioned it without one, left it out, or wrote no answer at all.

You will also see it sold as an "AI visibility tool" or an "AI visibility checker". The work it measures is optimising for AI answers (GEO): making your site the source an answer is built on, not one link in a list. How it differs from SEO is in SEO vs GEO vs AEO.

The choice matters more than it looks, because the tool's number is what you will act on for months: which page to edit, which competitor to track, whether the money you spent worked. A tool that measures a market other than yours, or judges from a single answer, hands you a confident wrong number, and that is worse than no number at all.

The eight questions below apply to any GEO tool you are considering. Here they are in four groups; each then gets its own section, and Tasadr's answers are gathered in one section near the end:

Six kinds come up when you search, and they do not all measure the same thing:

• SEO tools that have added an AI report. They know your site well in search results, and may add a glimpse of the engines' answers. Ask in which language they ask the engines, and on which engines.
• Global tools that track a brand in AI answers. They measure mentions of your brand on several engines, and many were built for English first, so ask whether they ask in Arabic and in your market's dialect.
• Customer-service chat tools. Bots that reply to your customers on your site. Useful, but they measure nothing of what ChatGPT or Google says about you.
• Google's and Bing's reports for site owners. Google Search Console shows how often your pages appeared in Google AI Overviews and Google AI Mode, and Bing Webmaster Tools shows their citations in Copilot. They are useful, but they do not measure ChatGPT or Gemini, and they do not show who was cited instead of you.
• Checking by hand. You ask the engines yourself and record the result. A good start that costs only your time; the method is in how to know if ChatGPT and Gemini cite your site.
• An Arabic-language platform for Arab markets. It asks in your customer's dialect and about their market, and reads Google's answers as they appear in their country.

It should ask in Arabic and in your customer's dialect, because the same question in English is answered from other sources about another market. Arabic answers are mostly built on Arabic pages, so a tool that translates your questions into English is measuring a different market and writing the result under your name.

Dialect is not a detail either. A customer in Jeddah and a customer in Kuwait word the same question differently, and the engine searches with the words it is given. Put three plain questions to the vendor.

In which language are my questions put to the engine? Can I write a question in my customers' dialect, the way they type it? Does the tool suggest Arabic questions, or translate English suggestions? If the reply is vague, ask to see one Arabic question exactly as it reached the engine.

The tool should show you the answer as it reaches your customer in their country, not a generic version: an answer in Riyadh can differ from one in Dubai for the same question. Engines often take the asker's country into account when they pick sources. A shop that delivers inside the Kingdom suits someone asking in Riyadh, and a shop in Dubai suits someone asking in the UAE, even when the question is worded identically.

This matters more in the Arab world than almost anywhere, because the markets sit side by side and share a language. A tool that asks from one country for every client, or gives you an average that blends markets, tells you something true about a place your customers do not buy from. Ask the vendor: from which country is my question asked, and can I choose the country for each site?

A number without text cannot be checked. Ask to see, under every verdict, the answer's excerpt as the engine wrote it, the links it cited, and the date. When a tool tells you that you are absent from an important question, you are entitled to read an excerpt of the answer you were absent from, to know who took your place, and when.

The text earns its keep twice. First, it proves the verdict: you may find the engine wrote your name differently, or cited a subpage of your site you had forgotten about. Second, it teaches you what your customer wants, because the engine restates the question and orders its answer by what it thinks matters. A tool that gives you a score and a chart and not one line of the answers is asking you to trust it without evidence.

A chat engine's answer changes from one time to the next, so a single snapshot may be a lucky answer. Ask for repeated measurement, and for the range the measurements moved within. Ask an engine the same question twice in an hour and you may read two wordings and different sources, because it searches afresh each time.

The risk cuts both ways. A tool that judges from one answer may congratulate you on a place you never held, or warn you of a loss that never happened, and you end up editing a page that was working. A tool that gives you a single number with no range does not tell you how sure it is. Ask the vendor: is each question measured repeatedly, or is my score built on one answer? Do I see the range when the measurements move? When does the tool call a change a real change?

A citation with a link is a visit; a mention without one is name recognition and nothing more. A tool that blends the two, or counts against your site an engine that wrote no answer, gives you a number you cannot act on.

A link brings the reader to you in one tap; a bare name tells them you exist and sends them back to searching. The third case many tools blur is the missing answer: some engines write no answer at all to some questions, and charging you for their silence lowers your score for a reason that has nothing to do with your site. Ask to see one verdict per answer, each with a fixed name, and to know which verdicts lift your score most.

They are two different Google surfaces, and one question can bring different sources in each, so ask for each to be measured under its own name. Google AI Overviews is the summary Google writes above its usual results page; Google AI Mode is a separate conversational mode in which Google writes a longer answer and cites its sources.

A tool that folds both into one line called "Google" hides a difference that may be the whole story: you can be cited in one and absent from the other, and each needs a different fix. The same goes for every other engine. Ask for the list of engines by name, and which of them are measured on the plan you will actually buy, not on the vendor's top plan.

Knowing who got ahead of you is half the answer. The other half is a comparison of your page with theirs, and an action plan ordered by importance that says what to change first. A list of names on its own is a report you read without knowing what to do with it.

A tool worth its price opens the cited competitor's page and compares it with yours on the same question: where the answer sits on their page, and what they said outright that you did not, such as a price, a city or a delivery time.

Then it turns what it found into an action plan that starts with the edit most likely to move you. Beware the tool that hands you raw answer text and leaves you to extract the meaning yourself: that is work landing on you, not a service you bought.

Ask for a fixed monthly price, published on a public page with tax included, and a free trial to test it before you pay. A tool that charges per question or per request turns every bill into a surprise and nudges you to measure fewer questions, which means measuring less than you need.

Ask as well: is the published price what I actually pay, or a starting point with fees per engine or per market on top? Can I cancel whenever I like? Are the plans and their limits written on a public page, or do I only learn them on a sales call?

Tracking mentions alone is not enough. Knowing your name came up in an answer does not tell you in which question, whether there was a link, or who came up instead of you in the questions you were absent from.

So ask a tracking tool to work by the question: for every question your customers ask, whether your site was cited with a link, your name mentioned without one, or you were absent, and who led instead of you.

An agency does the work on your site; a tool measures the result. A tool alone is enough if someone can write and edit your pages, and you need an agency or a consultant if nobody can do the work.

Who to choose, a platform, an agency or a consultant, and how to check them before you sign, is the subject of GEO agency or consultant: choosing in Saudi Arabia.

The best test of any tool is to measure your own questions with it, not the demo the vendor prepared. To see what a report worth asking for looks like, read a real report for an electronics store in Egypt. Five steps fit in one week:

• Write 5 questions your customers ask, the way they type them on their phones: in Arabic, in their dialect, with the city and the condition they care about, such as price or delivery.
• Add them to the tool and set your customer's country, then read the first measurement. Do you see an excerpt of every answer, the links the engine cited, and the date?
• Ask the engine yourself the same question from your phone in your country, and compare what you read with what the tool recorded. A word-for-word match is not the test, because answers change from day to day; the sources and the overall picture should be close.
• Open a cited competitor's page and read what the tool says about why it was chosen. Does it tell you something you can do tomorrow, or something general that fits any site?
• Read the pricing page again and work out what you would pay if you doubled your questions or added another site. If you cannot work it out alone, the price is not as published as it looks.

We gather our answers here in one place, so you can compare them with any tool's, question by question, and check every one yourself.

You can run this whole test in the 7-day free trial: one site, 10 questions and two engines, ChatGPT and Gemini, with no credit card. We run its two measurements ourselves: the first when setup is done and the second on day 5, once you prove the site is yours. What we measure and what we do not is gathered in frequently asked questions about measuring AI answers.

• Language. The suggested questions for each country of your site are written the way people there ask. You write questions in your customers' words, or we suggest them from your site's pages, or from real searches if you connect Search Console. Choosing the questions themselves is the subject of how to choose the questions you measure on.
• Your customer's country. You choose each site's country, and from Pro a site can hold more than one country, so each country's questions are written for its market and Google's answers are read as they appear there, not a generic version and not an average across markets. The list of markets holds Saudi Arabia, the UAE, Egypt and most other Arab countries. What the country means on Google in particular is covered in the guide to Google AI Overviews in Saudi Arabia.
• The answer's excerpt as evidence. Every verdict carries its evidence. From each answer we keep the excerpt that names you or your competitor, word for word as the engine wrote it, with the links it cited, the measurement date and the engine's name, and the scores for as long as the site is tracked. You read each question's latest excerpt and judge for yourself; what every measurement gives you is set out in how we measure AI answers.
• Repeated measurement. We measure every question again in every measurement, so you read your score across your measurements rather than from one, and when a measurement spans a range we show the range beside the score. An alert reaches you in your dashboard only when a change is confirmed: a competitor is not called new on a single answer, a citation is not called lost on a single answer, and a gain counts at once.
• Cited or mentioned. Every answer gets one of four outcomes with fixed names: Cited when your site's link is among the sources the engine relies on, whether or not it writes your name; Mentioned without a link when it writes your name alone; Absent when it writes an answer without you; and No answer when it writes nothing at all. A citation counts for more in the Tasadr score than a mention, and No answer is never counted against you.
• The engines by name. We measure 8 engines, each by name: ChatGPT, Google AI Overviews, Gemini, Perplexity, Google AI Mode, Copilot, Claude and Grok. The 7-day free trial measures two engines, ChatGPT and Gemini; Starter adds Perplexity and Google AI Overviews for 4 engines; Pro and Advanced add Google AI Mode for 5 engines; and Business measures up to 8 engines, chosen per client. Each engine is described in the eight AI engines we measure in Arabic.
• Why a competitor is cited. For every question you see the competitors cited instead of you, the why-they-lead analysis compares your page with the competitor's and says why the engine chose it, and you get an action plan ordered by priority. How many you get follows your plan: one why-they-lead analysis in the free trial, then 15 why-they-lead analyses a month on Starter, 50 why-they-lead analyses on Pro, 100 why-they-lead analyses on Advanced, and no limit on Business. How Tasadr (تصدَّر) compares your site with your competitors', question by question, is on the competitor comparison page.
• A fixed, published price. You start with a 7-day free trial on 10 questions and no credit card, then $49 a month on Starter, $129 a month on Pro, $249 a month on Advanced, and Business by quote, with a written quotation within one working day. Prices are in US dollars, paid in your local currency in Egypt, Saudi Arabia, the UAE, Qatar and Morocco, and final with tax included. A year costs ten months on Starter or Pro and eleven on Advanced, you can cancel at any time, and on a paid plan, if you need more questions than it holds, you add them in packs at a fixed monthly price. Every plan and its limits is on Tasadr's pricing.
• Tracking. We track your questions on the engines your plan covers, record one of the four outcomes for every answer, and keep a dated excerpt of it. An alert reaches your dashboard when a change is confirmed: a new competitor, a citation you lost, or a gain. We track what the answers say in your own questions, not every mention of your name on the web. How to read that tracking month after month is in how to measure the success of optimising for AI answers.
• For agencies. You can track your clients' sites with Tasadr: 3 sites on Pro, 5 sites on Advanced, and sites set to your agency's needs on Business. On Business, your agency's name and logo go on the reports your clients receive, and you download the weekly report as a PDF; on every plan you download your questions, your tasks and your monthly report as Excel. Your team also gets seats inside the account, by plan. The details are on Tasadr for agencies.

Six mistakes that cost more than the tool itself:

• Choosing on the number of engines alone. A long list is no use if the engine your customers ask is outside your plan.
• Judging from the demo. A demo is built on questions the vendor chose, in a market it knows well. Measure your own questions, in your own market.
• Trusting a number without text. A score you cannot open to read even one answer behind it is a score you can neither defend nor learn from.
• Measuring English for a market that asks in Arabic. You get correct numbers about a market that is not yours, and competitors who are not your competitors.
• Buying a dashboard instead of an action plan. Charts say where you stand; what moves you is what you change on your page this week.
• Waiting for a guarantee. Nobody can guarantee you a place in an answer, and anyone who promises one is selling a promise rather than a measurement. How ChatGPT chooses the sites it cites in Arabic explains why.