Choosing an AI visibility tracker, and whether you need one

Disclosure first: Dopbot sells an AI visibility tracker, so we have an interest in this question. The advice below is meant to hold even if you never buy ours, and the ten questions are answered for our own tracker as well, including where it falls short.

The decision comes down to repetition. A one-off look at how assistants answer shopping questions about your brand is cheap to do by hand. Doing it on a fixed set of prompts, every month, on more than one engine, against competitors, with ranges you can defend, is what turns a few hours into a job. This post puts numbers on that cost and lists what to ask a tool before you pay.

Do you need a tracker at all?

Start with the cost of doing it by hand. The arithmetic is answers times minutes per answer. Take the pattern of the index's monthly update as the unit: 5 question types, 2 engines and 4 repeats of each is 40 answers, the size of the monthly update for the 40 US categories that are still updated. The minutes per answer are an assumption: 3 for a fast, practiced person who records each answer in a spreadsheet, 5 for a careful one. Time yourself on three answers before relying on either figure.

Table: manual collection time under an assumed 3 or 5 minutes per answer, recording only; analysis, intervals and reporting come on top.

Manual is enough when you track one brand, look once or a few times a year, and are content with about 40 answers per question set. It is a good way to learn the method too; the free manual method walks through it. A tool starts to earn its fee when one of these holds:

• You need history: the same prompts, run the same way, month after month.
• You need more than 40 answers a month, across several question types or several brands.
• You track competitors and want the shared and exclusive answers counted for you.
• You report to clients, and the hours are billable time.

A break-even check keeps this honest. Divide the tool's monthly price by the hours it saves. The Brand plan is USD 79 a month, which at 10 hours saved means the tool pays off if an hour of your time is worth more than USD 7.90, and at 2 hours saved if it is worth more than USD 39.50. That holds only if the tool collects a comparable number of answers, which is question 3 below.

Which ten questions should you ask any tool?

For each, the answer for the Dopbot tracker follows. Where the public pages do not say, we say so.

1. Which engines, and through what? A tool built on developer APIs measures something different from one that records the consumer apps, and a tool that says "ChatGPT" may mean either. As of October 2026, Perplexity's own FAQ says its API and web interface outputs may differ, and the API versus app post covers the rest. Ours: OpenAI and Perplexity, each through its developer API with web search on. No consumer apps, and no Google.
2. With what settings? Location, web search, language, personalization. Ours: the index sets the market's country on each request and sends no account, history or memory. The tracker and the free snapshot use the United States.
3. How many answers sit behind each number? You need the denominator for every figure and every run. Ours: the public index category pages print each count as "k of n answers". The pricing page lists 25 prompts per brand, two engines and up to 5 competitors, and states how often the tracker runs; it does not state how many times each prompt is asked per run, so ask us before buying, as you would ask anyone.
4. Does it show a range, or say when a difference is too small to count? Ours: the category pages print counts and a plain warning that brands within a few answers cannot be told apart. They do not print intervals; we compute them in the walkthrough of a category page.
5. Who writes the prompts, can you edit them, and do they name brands? Prompts that name a brand change the answers. Ours: the plans build 25 buyer-style prompts per brand from your catalog, editable by you, per the pricing page.
6. How are brands found and merged? One brand under two names splits its count. Ours, for the index: an AI model reads each answer and lists the brands, a brand counts once per answer, and a hand-kept list merges known name variants. A variant nobody has reviewed is not merged.
7. What does a source list contain? Pages an answer links and pages a search retrieved are different things, and adding them together misleads. Ours: we keep the two engines apart, and say that OpenAI's list is the pages its answer links while Perplexity's is the pages its search retrieved.
8. Can you get the data out? Ours: the index's public CSV files hold counts per brand and category, that is, appearances out of samples. They are not the raw answers. For the tracker, the pricing page does not say what can be exported.
9. How does it decide that a change is real? A tool that shows arrows from one run to the next needs a rule. Ours: the index shows no change over time, because the October 2026 edition and the monthly updates are not like for like; its methodology says a change will be listed only after an exact test gives p below 0.01, and that test is not built yet.
10. What are the price and the terms? Ours: Brand USD 79 a month for 1 brand, Agency USD 199 a month for 5 brands, billed monthly through a payment provider, no contract, cancel anytime. What happens to your data after you cancel is not stated on the pricing page.

Where does Dopbot's tracker fall short?

Plainly, so you can weigh it:

• Two engines. OpenAI and Perplexity only. Other assistants and Google's AI features are not measured.
• API sampling. The answers come from developer APIs, not from the consumer apps, and they carry no personalization. OpenAI is sampled through its API with web search (gpt-6-luna) and Perplexity through its Agent API with the Sonar model (perplexity/sonar) with web search on, so neither is a recording of what the apps show.
• A public index that updates monthly, with coarse categories. The tracker's own run schedule is on the pricing page. The October 2026 edition has about 10 answers per category; 40 US categories now have 40 a month, and the other 120 categories, including all of Mexico and Canada, are frozen at October 2026.
• United States only for the tracker and snapshot.
• No guarantees. The pricing page says Dopbot does not promise more recommendations, and that is the correct position for any tool.

Which answers should make you walk away?

• A promise that the tool will get you recommended. Google's own guidance on hiring an SEO says no one can guarantee a #1 ranking, and that you should find someone else if a provider guarantees first place; nobody controls what an assistant says.
• A single "visibility score" with no definition, no denominator and no list of prompts.
• Engines merged into one number, or an "app" result that is really an API call.
• A percentage change between two runs with no sample size.
• Rankings of brands that are a few answers apart.
• A refusal to say how many answers each figure rests on.

Frequently asked questions

Can I start by hand and move to a tool later?

Yes, and it is a good order. Keep your prompt list, your recording rules and your first results in a spreadsheet. A tool you choose later can then run the same prompts, so your baseline stays comparable. Compare the tool's first results against your manual ones; large gaps mean a difference in setup, such as the engine, the location or the way brands are counted.

Is the free option enough?

For a first look, usually. The public index pages cover 160 categories: 120 are frozen at the October 2026 edition and 40 US categories are updated monthly. A free snapshot gives a one-off check for your own brand. Neither gives you a fixed prompt set run over time with your competitors, which is what a paid tracker is for.