AI Visibility Tracker or a Spreadsheet? Seven Questions and the Real Prices

You can measure whether assistants name your brand with a free sheet and an hour a month, or you can pay a vendor between thirty and five hundred dollars a month to do it on a schedule. This piece settles which one you need now, and gives you the trial that proves it. No vendor is ranked here. We sell content software, so our interest is in you having criteria rather than a shortlist.

The verdict in forty words

Log twenty to thirty real buyer questions by hand for one month first. Write down what you wanted and could not get. Buy only when that list names scale, history or competitors, and only after a trial run on your own questions.

Side by side, where the two actually differ

Dimension	A sheet you keep	A subscription
What it costs	Roughly an hour a month for twenty-five questions across two assistants	A monthly fee, tiered by question count
Ceiling	A few dozen questions before it stops being realistic	Hundreds, on a schedule
Back history	Only what you wrote down	Collected from day one
Repeat runs	Yours to do, by hand	Usually automatic
Rivals	One at a time, manually	Normally built in
What you learn	Every answer, in full, including the wording	A rolled-up figure
How it fails	You log inconsistently and the series breaks	You believe a score you cannot open up

Stay manual while these four things are true

• This is new to you. Reading the raw replies is how you find out what vocabulary assistants use for your category, which no dashboard will show you. Our four-step logging method is the whole setup.
• Your question list is short. Under about thirty questions, doing it by hand is an afternoon, not a project.
• You still do not know what to score. The AirOps guide to citation tracking grades each reply on whether the brand is named, whether the description is accurate, how it is framed, and what kind of citation appeared. Every one of those is a human judgement you can make faster than you can configure.
• The budget is small. A year of an entry subscription is a freelance writer for a month. One of those changes whether you get cited.

Buy the subscription once these are true

• Scale has become the bottleneck. A few hundred questions across several assistants and two or three regions cannot be done by hand, at any discipline level.
• You report monthly, upward. Stored history and a chart that draws itself genuinely save hours every month.
• You need rival numbers. Share of voice only means something if every competitor is measured the same way, on the same day. By hand, they are not.
• You run this for clients. Agencies need the same report, reproducible, across accounts.

Before comparing two vendors, check that they define the words the same way. A brand being named in an answer and your page being used as a source are separate events with separate fixes, which is the point of splitting those two metrics and of pinning down what mention, citation and link each mean.

What an entry tier actually buys, at current list prices

Published pricing is where the comparison gets concrete. Otterly's own page lists Lite at $29 a month for fifteen tracked prompts across ChatGPT, Google AI Overviews, Perplexity and Copilot, then Standard at $189 for a hundred prompts and Premium at $489 for four hundred, with Claude, Gemini and Google AI Mode sold as paid add-ons on top of every tier. Respona's roundup corroborates the Lite figure and puts Profound's entry plan at $99 a month for ChatGPT alone, with multi-engine tracking starting at $399. Peec takes the other approach: its plan names are public and its prices are not shown on the pricing page at all.

Two implications worth sitting with. The assistant your buyers actually use may be an add-on rather than a feature, so the list price is not the price. And prompt allowances are the real unit of cost: adding a hundred more prompts to a mid tier is itself a line item on Otterly's price list.

Seven questions for a trial, in the order that saves money

Surfer's buying guide sets out what a capable tracker covers, including coverage across several assistants, scoring with history, competitor tracking and traffic estimates (Surfer). Turned into questions a vendor has to answer during a trial, cheapest first:

1. How many prompts does my plan include, and what does the engine I care about cost on top? Settle this before the demo. It decides whether the tool is affordable at the size you need.
2. How many runs make up one score? One run per prompt per day is a sample of one, and answers move between identical runs.
3. Can I open the raw answer behind any data point? If not, the score is unauditable and you will eventually have to defend it.
4. What is the formula? Ask for it in writing. A refusal is an answer.
5. Can I load my own prompt list, unedited? A tool that generates its own questions is measuring its idea of your market.
6. Where does the traffic estimate come from? Surfer's guide is blunt that assistant traffic cannot be forecast reliably from outside, so any number here is modelled. Ask what the model is.
7. Can I export the history if I leave? Stored history you cannot take with you is the vendor's asset, not yours.

The cheapest plan is a trap, and the sheet is not the alternative to it

Here is the position: the entry tier of most of these tools is priced to be bought and is too small to be useful, and the honest comparison is a manual sheet against a mid tier. The evidence is in the allowances. Fifteen tracked prompts, which is what Otterly's Lite plan includes, is half the question set a credible baseline needs, and it buys one run per prompt per schedule. Respona makes the related point that a visibility percentage is relative to the prompts you chose, not absolute, so a small prompt set does not just measure less, it measures something narrower and flatters or punishes you by accident. At that size you are paying a subscription for a worse version of what a free sheet already gives you, minus the raw answers.

The fair objection: an entry plan is a cheap way to learn the category, and automation beats a sheet nobody keeps filling in. Both are real. Discipline is the single biggest failure mode of manual tracking, and a tool that runs itself on a Tuesday will beat a sheet you abandon in March.

So one condition decides it. If you have tried the manual month and the log has gaps, buy the cheapest tier that holds your actual prompt list, not the cheapest tier on the page. The thing to refuse is the entry plan bought instead of the manual month, because then you have neither the habit nor the coverage, and a dashboard that reports confidently on fifteen questions.

Four weeks that settle it

1. Week one. Build the manual baseline: twenty-five buyer questions, two assistants, every answer saved. Freeze it.
2. Week two. Put the identical list into a vendor trial and compare question by question, not in aggregate.
3. Week three. Take the five biggest disagreements to the vendor and ask for an explanation of each. Re-run those five by hand the same day.
4. Week four. Name what the tool showed you that your sheet could not, usually competitor trend lines. Price that against the annual cost.

Broad agreement across the list earns the tool some trust. Wide disagreement needs explaining before you rely on the score. A vendor who welcomes this comparison is telling you something, and so is one who steers you away from it.

Earn the subscription with one month of manual logging

Also check your own analytics before you buy anything, because mentions and visits are not the same signal. In a March 2025 browsing panel, Pew Research Center recorded a click on a link inside a Google AI summary during 1% of visits where a summary appeared, so a tracker can show you rising presence while referral numbers stay flat, and both readings can be correct. Run the sheet for a month, write the list of what it could not tell you, and take that list to two vendors. If the sheet already named the pages to fix, spend the subscription on fixing them instead.