How to choose an AI visibility
audit provider.
A practical guide for consumer brands evaluating who should measure and fix their AI visibility. Written by one of the options, which you should factor in. Every competitor price below is quoted from that company's own published pages as of August 2026.
There are three genuinely different things being sold under similar language, and most confusion in this market comes from comparing across categories as though they were substitutes. Sort out which category you need first, then compare within it.
The three things you might be buying
A one-time audit
Point-in-time diagnosis
Someone runs a prompt set for your category, scores where you stand, checks your site's technical readiness, and hands you a prioritized fix list. It ends. You own the output.
Buy this when you do not yet know why you are invisible, or you know and need the fixes sequenced. Typical range: low four figures. Shelf Help AI is $1,950.
A monitoring platform
Continuous measurement
A subscription dashboard that tracks your citation rate, sentiment, and competitor share across assistants over time. It reports. It does not remediate, and it does not interpret.
Buy this when you know you have a problem, want the trend line, and have someone in house to act on it. Published entry prices: roughly $29 to $295 a month, with useful tiers higher.
A retained agency
Ongoing outsourced execution
A team works your AI visibility continuously: strategy, content, technical implementation, measurement. You are buying capacity and judgment on an ongoing basis.
Buy this when you have budget and would rather not build the capability in house. Published entry points: iPullRank "starting at $15,000/month"; NoGood "average retainer is above $20,000/month."
The most common mistake we see is a brand buying category 2 when it needed category 1. A dashboard showing you are cited in 18% of category prompts is not actionable if nobody at the company knows whether 18% is good, what is causing it, or which of forty possible fixes moves it. Measurement without a baseline diagnosis is an expensive way to stay confused.
What a good AI visibility audit should measure
Use this as a checklist when you evaluate anyone, including us. If a provider cannot answer these, keep looking.
On the prompt side
- Sample size. How many prompts? A single AI answer is close to a coin flip. Fewer than twenty and you are reading noise
- Prompt realism. Are they buyer-style questions a shopper would type, or branded searches? A prompt naming your brand teaches you nothing about whether you get recommended
- Model coverage. Which assistants, and in which modes? ChatGPT, Claude, Perplexity, Gemini and Google AI Overviews disagree with each other constantly
- Session hygiene. Are prompts run cold, with no chat history or custom instructions? If not, the results reflect the tester, not the model
- What is scored. Citation rate alone is not enough. Ask for position within the answer, factual accuracy of what is said about you, and competitor share
- Evidence retention. Can you read the actual answers? Every claim should trace back to something you can verify
On the technical side
- Bot governance. Do they check that GPTBot, ClaudeBot and PerplexityBot are actually admitted, including at the CDN and WAF layer, not just in robots.txt
- Structured data depth. Not just "do you have schema" but whether Product markup carries real attributes and whether FAQPage matches visible content
- Commerce data. Most SEO-derived audits skip this entirely. Ask whether they check attribute depth, pricing and availability accuracy, and return policy clarity
- Entity consistency. Does your brand resolve consistently across your site, retailer listings, social and third-party sources
- Prioritization. Does the output rank findings by impact and effort, or is it a flat export of everything that failed
- Ownership. Do you keep the roadmap and can your own team execute it without them
The monitoring platforms, compared
If category 2 is what you need, these are the options that came up most often in our research. All published prices verified August 2026. We do not sell any of these and receive nothing if you buy them.
| Platform | Published entry price | Worth knowing |
|---|---|---|
| AthenaHQ | Free tier with 300 credits; Starter $295/mo | The most developed consumer-goods product surface of the group, with a dedicated CPG practice. Prices credits as "1 credit = 1 AI response," which is unusually clear. Note that Starter is single-region and single-language, and API access plus extra credits are add-ons priced on request, so $295 is a floor rather than a total |
| Scrunch AI | $250/mo billed annually, or $300 month to month | The most transparent pricing we found: month-to-month option published, seat add-ons published at $25, and a seven-day trial with no credit card. Also covers Meta AI, which the others generally do not |
| Profound | $99/mo Starter, $399/mo Growth, both billed yearly | Starter tracks ChatGPT only; Growth covers three assistants. ChatGPT Shopping tracking, the most relevant feature for a consumer brand, is Enterprise-only and Enterprise pricing is not published. Their FAQ notes teams over three people are "better suited for a custom enterprise plan" |
| Otterly.ai | $29/mo Lite, $189 Standard, $489 Premium; Enterprise "starting from $1,000/month" | Lowest sticker price and the only one publishing an enterprise floor. Compare total cost carefully: Claude, Gemini and Google AI Mode are paid add-ons at every tier, so full model coverage costs considerably more than the headline. Monthly cancellation is explicit |
Figures read from each company's own published pricing pages in August 2026 and quoted rather than estimated. None of the four publishes an Enterprise price except Otterly.ai. Verify before purchasing; pricing in this category is changing quickly.
Questions worth asking any provider
"How many prompts, and can I see them?"
A provider who will not show you the prompt set is asking you to trust a black box. The prompts are the methodology. If they are generic rather than built for your category and need states, the results will be generic too.
"What happens when the engagement ends?"
Ask specifically what you keep. A roadmap your team can execute is a durable asset. Access to a dashboard that switches off when you stop paying is not the same thing, and both are legitimate purchases as long as you know which one you are making.
"Would you tell me not to buy this?"
Worth asking directly. A provider whose revenue depends on a subscription is structurally unable to tell you that you do not need one. Listen for whether they will name a situation where a competitor or a free approach serves you better.
"What do you think is actually causing it?"
The answer separates diagnosis from measurement. "We will track it and see" is a tool answer. A useful provider should be able to look at your site for twenty minutes and form a hypothesis before you have paid them anything.
Where Shelf Help AI fits, and where it does not
A good fit
Consumer brands roughly $5M to $500M in revenue that want to know where they stand and what to fix, on a fixed price with a defined end, and want the roadmap to belong to their own team. $1,950 for the audit, about a week, no call required. Optional monthly tracking at $295 afterward if the trend line is worth having.
See the audit →Not a fit
If you need a continuous dashboard as the primary deliverable, buy a platform from the table above; several are excellent and we will happily say which suits your setup. If you want a team executing continuously, the retained agencies are built for that and we are not. And if you have not yet checked whether you have a problem, do the free version first and possibly save yourself the whole exercise.
Start free →Common questions
How do I choose an AI visibility audit provider?
First decide which of three things you are buying: a one-time audit that diagnoses and hands you a fix roadmap, a monitoring platform that tracks visibility continuously, or a retained agency that executes on an ongoing basis. Then evaluate within that category. For an audit, ask how many prompts are run and whether you can see them, which assistants are covered and in what modes, whether prompts run cold in fresh sessions, what is scored beyond citation rate, whether evidence is retained, and whether the technical review covers bot governance and commerce data rather than only classic SEO checks. Ask what you keep when the engagement ends.
How much should an AI visibility audit cost?
For a mid-market consumer brand, a fixed-scope audit sits in the low four figures; Shelf Help AI charges $1,950. Monitoring platforms publish entry prices from roughly $29 to $295 per month, with genuinely useful tiers generally higher and some assistants sold as add-ons. Retained agency programs are an order of magnitude above both: iPullRank publishes "starting at $15,000/month" and NoGood states its average retainer is "above $20,000/month." Those are different products rather than the same product at different prices, so compare within a category rather than across.
What is the difference between an AI visibility audit and an AI visibility tool?
An audit is a point-in-time diagnosis performed and interpreted by a person, ending in a prioritized fix roadmap you own. A tool is a subscription that measures continuously and reports, leaving interpretation and remediation to you. They answer different questions: an audit answers why you are invisible and what to do about it, while a tool answers whether the number is moving. Brands that buy the tool first frequently discover they cannot act on it, which is why the diagnosis usually belongs first.
How many prompts should an AI visibility audit run?
Twenty to thirty per category, minimum, run cold in fresh sessions with no chat history. A single AI answer is close to a coin flip: the same prompt run twice can name different brands. Below roughly twenty prompts, citation rate and competitor share move around enough that you are reading noise rather than a position. Any provider running five or ten prompts and presenting the result as a measurement is overstating what the sample supports.
Do I need both an AI visibility audit and a monitoring subscription?
Not necessarily, and it is worth deciding deliberately. Many brands need only the audit: they get the roadmap, execute the fixes, and re-check informally later. Tracking earns its cost when a brand has made real changes and needs to know whether they worked, or operates in a fast-moving category where competitor movement matters month to month. Buying continuous measurement before knowing what the number should be for your category is the sequence most likely to waste money.
Can I do an AI visibility audit myself?
Yes, at least partly, and you should do the free parts before hiring anyone. Check your robots.txt for GPTBot and ClaudeBot. Run a schema validator against your product pages. Write twenty buyer-style prompts for your category and run them cold across ChatGPT, Claude, Perplexity and Gemini, logging every answer. That afternoon of work will tell you whether you have a problem. What it will not do is tell you which of forty possible causes explains it or which three fixes matter most, which is where outside help starts earning its fee.
Start with the free diagnostic.
Whoever you end up hiring, knowing whether you have a problem should not cost anything. The free AI Shelf Snapshot runs five buyer-style prompts for your category and returns a one-page scorecard in two business days.
Get Your Free SnapshotComparing specific firms? See the head-to-head comparison →