Most AI visibility tools measure the same thing and disagree about the answer. That is not a bug in one of them, it is the nature of the metric: ask an assistant the same question twice and it rarely returns the same list of brands. So the useful question when comparing these tools is not who has the most engines. It is whether the number they hand you comes with any indication of how much to trust it.

Here is the comparison at a glance, then the detail.

Tool	Best for	Starting price	Engines	Shows uncertainty?
LadderFox	Solo founders and small SaaS teams	$0 free, paid from $19/mo	3 on paid plans, included in the price	Yes, 95% interval and sample size on every score
Otterly	Cheapest daily tracking	$29/mo	4 included, 3 more sold separately	No
PromptWatch	Engine breadth plus crawler analytics	$95/mo, free tier	More than 10, included	No
Peec AI	Mid market teams and agencies	$95/mo	6 included, more sold separately	No
Profound	Enterprise procurement	On request	Broad	No
Writesonic	Visibility inside a content suite	Suite pricing	Several	No

Prices verified 31 August 2026 from vendor pricing pages, and prices in this category move. Note the fourth column before the second one: three of these vendors quote an entry price and then sell extra AI engines separately at $9 to $149 per month per engine, which can double the sticker before the tool covers what the homepage promised.

The axis that actually separates them

SparkToro ran 2,961 AI queries with 600 volunteers and found that two identical runs return the same brand list less than one time in a hundred. Kevin Indig found only 2.2% of ChatGPT citations survive three runs of the same prompt, and that engines replace between 56% and 74% of their cited sources week to week.

That has an uncomfortable consequence. A visibility score of 34 means very little on its own. Ran again an hour later it might read 21, and nothing about your brand changed. Any tool that draws a single line through those readings is inviting you to see movement where there is noise, and any tool that emails you about a 13 point drop is wasting your afternoon.

So when you compare these products, look for three things:

How many times is each question asked? Once is a coin flip. A reviewer of one tool in this list put the industry state plainly: how many runs sit behind a reported number is "not documented, the same gap almost every vendor here has".
Is there a range around the score? A confidence interval is the difference between "you appear in 20% of answers" and "somewhere between 6% and 51%, and we only asked ten times".
Are branded and unbranded questions separated? Asking "is Acme any good" and asking "best tools for X" measure completely different things. Mixing them inflates every score, and it inflates yours most.

Almost nobody in this category does all three. That is the honest state of the market rather than a sales point.

LadderFox

Free for one check, then $19, $79 and $199 a month. Every plan price includes its engines.

Every score carries a 95% confidence interval and the sample size behind it. A change is only reported as a change when the intervals stop overlapping, which means the product will tell you that a 23 point move is statistically indistinguishable from nothing. It also splits unaided from aided, so you can see the common and depressing pattern where an engine describes your product accurately once someone names it and never reaches for it otherwise. As far as we can verify from vendor documentation, it is the only tool in this comparison under $189 that prints the uncertainty behind its number.

It also joins two halves that are usually sold separately: a server side collector shows which AI crawlers read which pages, and that gets lined up against which of your pages the engines went on to cite. That produces a list most tools cannot produce, which is the pages that get read constantly and quoted never.

Recommendations carry a source. The action plan can only suggest tactics that cite a published study from a vetted registry, so you can check the reasoning rather than take it.

Where LadderFox is the wrong choice. It covers five engines in total where PromptWatch covers more than ten, so if you need Grok and DeepSeek specifically, this is not the tool. There is no public API and no MCP server yet. There is no agency seat management, so if you are reporting on twenty client brands you will feel it. It is the youngest product here with the shortest track record. And it deliberately does not show estimated prompt volumes, which some buyers genuinely want, for the reason in the last section.

Otterly

$29 a month for the Lite plan, then a jump to $189 and $489. Seven day trial, no card.

The cheapest way into daily tracking, and the most reviewed budget option in the category. Lite covers 15 prompts across four engines: ChatGPT, Google AI Overviews, Perplexity and Copilot. Every plan tracks daily and seats are not limited, which is genuinely generous at this price.

Two things to check before buying on the sticker. First, Claude, Gemini and Google AI Mode are not in the base price: they are sold separately per engine, and one reviewer computed that full engine coverage takes the $29 plan to roughly $76 a month. Second, there is nothing between $29 and $189, so the moment 15 prompts stop being enough the price multiplies by six and a half.

On methodology, Otterly documents its collection approach, querying the engines as a neutral user rather than through APIs, but publishes no runs per prompt and no interval. Their own blog acknowledges that identical prompts produce different answers and answers the problem with reassurance rather than a published stability number.

Pick Otterly if you want the cheapest daily tracking across the big four engines and you are comfortable reading trends without error bars.

PromptWatch

Free Explore tier, then $95, $245 and $579 a month.

The broadest engine coverage of the self serve options, more than ten including ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, with no per engine charges. Its strongest feature is agent analytics: it detects AI crawlers hitting your pages server side, so you can see whether the engines are reading you at all rather than only what they say. It also tracks citations from Reddit, YouTube and news, has content gap detection, and ships an API and an MCP server.

What is missing is the methodology. Nothing on their site or in the hands on reviews states runs per prompt, sample sizes or confidence intervals. One reviewer ran twenty prompts across ten engines for a week, logged 662 responses, watched citation rank move around, and the product never quantified that movement. It also displays estimated prompt volume, which the same reviewer described as directional rather than an exact figure.

Pick PromptWatch if engine breadth and crawler visibility matter more to you than knowing the error bars, or if you want an API today.

Peec AI

$95 a month for Starter with 50 prompts, then $245 and $495, plus a separate agency line. Extra AI models beyond the six default engines are sold separately by tier.

The fastest growing product in the category and, by most reviews, the cleanest interface. Peec classifies every prompt as branded or unbranded, publishes real research, and lets you drill from any metric down to the raw answer behind it. A vendor that publishes its numbers is easier to trust than one that does not.

The gap sits between their research and their product. Peec's own published studies state that every prompt was run multiple times "to account for the inherent variance of LLM responses". The product, by its own documentation, measures one run per prompt per engine per day and reports the result as a point value with no interval. The research team samples; the customer sees a single coin flip per day.

Pick Peec if you want the branded split, agency workspaces and a polished European vendor at a mid market price.

Profound

Enterprise, priced on request.

The largest and best funded company in the category. If you have a procurement process, a security review and a budget that starts in four figures a month, Profound is the safe shortlist entry. It is not the tool you evaluate on a Tuesday afternoon with a credit card.

Writesonic

AI visibility inside a broader content and SEO suite.

Its distinguishing feature is the verify loop: after you act on a recommendation it measures again at set intervals so you can see whether anything moved. Useful if you want visibility tracking bundled with the writing tools rather than as a separate subscription and a separate login.

On prompt volumes

Several tools in this category show a search volume style number next to each prompt. Treat those with care. Nobody outside the model providers has access to real prompt frequency, and independent testing has found published figures off by very large multiples. LadderFox does not show them at all, which is a deliberate gap rather than a missing feature, and it is worth asking any vendor where their number comes from before you plan a quarter around it.

How to choose in an afternoon

Run the same three prompts your buyers would actually type through two or three of these tools on the free or trial tier. Then ask each one the same question: how many times did you ask, and how sure are you? The answers will separate these products faster than any feature table, including this one.

Questions people ask

What is the cheapest AI visibility tool? LadderFox Signal at $19 a month is the cheapest tracking plan in this comparison. Otterly Lite at $29 is the cheapest with daily checks, though full engine coverage there costs extra. Both LadderFox and PromptWatch have free tiers with no card.

Which AI visibility tools show a confidence interval? Of the six compared here, only LadderFox publishes a confidence interval and sample size with every score. None of the other five documents how many runs sit behind a reported number.

Which tools include all their engines in the price? LadderFox and PromptWatch include their engines in the plan price. Otterly and Peec sell extra engines separately at $9 to $149 per month per engine.

How many times should a prompt be asked before trusting the score? Research on answer variance found identical prompts agree on the brand list less than one time in a hundred, so a single run is a coin flip. What makes a score trustworthy is distinct questions with a reported interval, not one question asked once and drawn as a line.

What is the difference between aided and unaided visibility? Unaided means the engine names your brand when nobody mentioned it, which is what wins customers. Aided means the engine can describe you once someone types your name. Tools that mix the two inflate every score. LadderFox and Peec separate them; most others do not.

Is a more expensive plan a more accurate measurement? Usually not. Higher tiers in this category buy more prompts, more brands and faster cadence, not better statistics. If accuracy is the concern, ask about runs and intervals, not tiers.

Method

Prices and features here were checked on 31 August 2026 from vendor pricing pages and hands on reviews, and prices in this category move. The variance figures come from SparkToro's volunteer study, Kevin Indig's citation persistence work and Ahrefs' analysis of 26,283 cited pages. Quotes about undocumented methodology come from independent tool reviews. LadderFox is our own product, which is why its section is the only one with a list of reasons not to buy it, and why it sits first in the table: not because we ranked it, but because this page is ours and you deserve to know that before you weigh it.