"AI SEO tool" is used for two different things. One group of tools does SEO work with AI: they write copy and suggest keywords. The other group measures whether your brand shows up in the answers of AI assistants. This article is about the second group: tools that measure visibility in ChatGPT, Gemini and Perplexity.
These tools share the same core: they ask buyer questions to AI engines at regular intervals and record how often your brand is mentioned, how competitors are doing and which sources are shown. The differences are in the details, and the details decide whether the measurement is reliable. Ask these seven questions when you choose.
1. Who decides the questions?
The value of the measurement depends on the question list. A high score measured with the wrong questions will mislead you.
What to check:
- Does the tool suggest questions, or do you have to write the list from scratch?
- Does your brand name appear in the suggested questions? If it does, the tool measures awareness, not visibility. For a question like "Is company X any good?", the engine will mention you anyway.
- Can you edit and approve the list? The measurement should run on questions approved by someone who knows the business.
2. Are the engines measured separately?
ChatGPT, Gemini and Perplexity look at different sources and give different answers. You can be strong in one and invisible in another. A single score with no breakdown by engine hides this difference.
Also ask this: what happens when an engine does not return an answer? Is that result recorded as missing, or is the gap filled with another engine's answer? The second option quietly breaks the measurement.
3. Does it rely on a single answer or on repeated measurement?
AI answers are volatile. In research SparkToro ran in late 2025, the chance of the same list of brands appearing in two answers to the same question was below one percent. The researchers concluded that individual answers and the question "what rank am I in AI?" are misleading, while the mention rate measured across many questions and repeats is a reasonable indicator.
So make sure the tool measures enough questions at regular intervals and shows the result as a trend over time. Be cautious with tools that present rank as the main metric. Rank is supporting information.
4. Which metrics does it give?
A useful measurement should show at least the following:
- The share of answers that mention your brand.
- Your share of all brand mentions in the answers (share of voice).
- How your competitors do on the same questions, and their trend over time.
- The raw answer for each question and engine: what the engine actually wrote.
The last item matters. If you cannot see the sentence behind the score, you cannot verify the result.
5. Does it show the sources?
The measurement tells you what is happening. The sources tell you why. AI answers are often built from other pages. In an Ahrefs study on ChatGPT, 43.8 percent of the cited links were "best X" blog lists.
Check whether the tool answers these questions:
- Which domains do the engines cite as sources?
- Which of these are list pages?
- Which pages include your competitors but not you?
- What should you do, and who should you write to, to get on these pages?
Be careful about who you write to as well: tools that generate guessed email addresses can lead you to the wrong person. An approach that shows the contact options actually found on the page is safer.
6. What are the recommendations based on?
Many tools give recommendations next to the measurement: change this on your page, add this section. The main risk here is that the AI makes up information that is not on your site: a customer count that does not exist, a wrong founding year, a service you do not offer.
Questions to ask:
- Are the recommendations prepared only from the information on your site?
- Does the text the tool suggests changing actually appear on your page?
- Are numbers you do not have removed, or are they added to the text?
Publishing a recommendation that contains made-up information costs you trust instead of gaining visibility.
7. Does it run technical checks?
If AI bots cannot access your site, all the other work is wasted. According to OpenAI's documentation, sites that block OAI-SearchBot are not shown in ChatGPT search answers. At a minimum, the tool should check robots.txt, whether AI bots are blocked, the sitemap and basic page tags.
What should you look at when choosing a price and plan?
When you compare prices, look at these points next to the monthly fee:
- How many questions are measured, and how many engines are they asked to?
- How often is the measurement run? AI answers change slowly. For most brands, weekly measurement is enough.
- Is the work that follows the measurement (source analysis, page recommendations) included in the plan?
- Is there a free trial, and does it ask for card details?
- Is the price in Turkish lira (TL) or a foreign currency? For teams in Turkey, this affects budget planning.
What should you test during the trial?
Most tools offer a free trial. Do not spend the trial just looking at screens. Check these:
- Read the suggested questions. Do your buyers really ask like this? Does your brand name appear in the questions?
- Ask one question in ChatGPT yourself and compare it with the answer the tool shows. It will not be identical, because answers change. But the overall picture should be similar.
- Search for a competitor's name. Does the tool show the answers where your competitor is mentioned?
- Open a page from the sources list. Is it really a page about your category?
- Compare a text recommendation with your page. Is every piece of information in the recommendation on your site?
These five checks show within a few hours whether the tool gives reliable results in your category.
How do you avoid tools that promise guarantees?
A simple rule: a tool or service that says "we will get ChatGPT to recommend you" is not promising something realistic. AI answers depend on how the engines work and on third-party sources. An honest tool measures, shows the reasons and tells you the steps that raise your chances of being recommended.
How does AIShortlist answer these questions?
Let's describe our own product by the same criteria. AIShortlist suggests questions from your site, your brand name does not appear in the suggestions, and measurement does not start until you approve the list. ChatGPT, Gemini and Perplexity are measured separately. If one does not work, the result is recorded as missing and is not filled with another engine. Every answer that names you shows the sentence where your brand appears. Each question is asked to each engine once per measurement; the repetition comes from the weekly runs. The Sources screen shows list pages and the places where you are missing, and the action plan uses only the contact options found on the page. Page text recommendations are prepared only with information that appears on your site, and numbers that are not on your site are removed. The technical site check is included in every plan, including the free trial.
Our limits, too: today only ChatGPT, Gemini and Perplexity are measured; Google AI Mode and AI Overviews are not, and question language is English or Turkish. How the measurement works is on the methodology page, and what the output looks like is in the sample report.
Paid plans start at $79 a month, and measurement runs every week. Before you decide, you can run a 10-question measurement with the free trial without giving card details. Whichever tool you choose, the seven questions above give you a solid list for making the right decision.