AI visibility tools, and what they will not tell you

Should you buy any of them?

Honestly: probably not yet, and not because they are bad. Because of what the evidence says they can measure. A tracker samples a handful of prompts on a schedule. Real users type millions of variants, and the same prompt returns a different list of brands more than 99 times out of 100. A number built on that foundation moves for reasons that have nothing to do with your marketing.

Benjamin Houy built one of these tools, ran it for seven months, and shut it down. His conclusion, which cost him a company to reach:

Customers were churning because the product didn't change what they needed to do. There's no secret GEO strategy. AI models reward the same fundamentals that already drive SEO and PR.

If you need a number to report upward, buy the cheapest one that names its methodology and treat it as a weather vane, not a speedometer. If you want to be cited more, the evidence points at the unglamorous things: be genuinely worth citing, get mentioned in places that already are, and make your pages easy to quote. None of that requires a subscription.

How to evaluate a tool if you choose to buy one anyway

If your organisation has decided that a top-line number is non-negotiable (because it reports upward, or because it helps calibrate attention against competitors), then choosing between them matters. Here is what to ask before you sign a contract:

• Does it name its methodology? If it does not say which models it tests, how often, or how many prompts it runs, you are paying for a black box. Gumshoe and Nightwatch name theirs; Profound does not.
• Does it test each engine separately or blend them? A blended "AI visibility score" is mathematically averaging three near-disjoint populations (Kevin Indig found only 2.37% of cited URLs appear across ChatGPT, Perplexity and Google AI Overviews for the same prompt). Tracking them separately is more honest and more useful: a change in your Perplexity citations tells you something; a move in a blended score tells you almost nothing.
• Does the sample include real competitors? Many trackers sample brand-recommendation prompts, which are high-variance and personalisation-heavy. Sample prompts that actually match search intent in your category. A tool that tests fifty randomised queries is cheaper but useless; one that tests twenty you selected is less flashy but far more actionable.
• Can you audit the prompts manually? Run the same prompts yourself on the same engines the tool uses. If your results diverge from the tool's, that is immediate evidence that the measurement is off.
• What is the raw data retention? A tool that deletes historical runs after 90 days is cheaper to run, but it prevents you from seeing trends. Better tools let you export the raw prompts and responses, so you own your own data.

When tracking makes sense

Tools are most useful not as a marketing number but as a research instrument. If you are deciding whether to redirect resources toward AI visibility, a baseline measurement helps you set expectations. If you are testing a hypothesis ("if we get featured in TechCrunch, our Perplexity citations will rise"), tracking across the experiment period is cheaper than waiting six months to guess. If you are in a category where AI assistants are the majority search method (technical documentation, research, open-source projects), then knowing which queries trigger your competitors' citations and whether you appear for them is legitimate intelligence.

What is not useful: buying a tool to track a "GEO score" every month and reporting it as a KPI. The measurement is too noisy, the signal too weak, and the cost unjustified by the return. The fundamental levers do not change. Be genuinely worth citing. Get mentioned in places that already drive AI citations (Reddit, Stack Overflow, industry publications, Wikipedia, licensed data partnerships). Make your answers quotable. And measure outcomes you can act on: traffic, conversions, branded search volume. A tool that does those things better is a better tool.

What the data shows about tool accuracy

Tools vary wildly in their stability across runs. Nightwatch publishes confidence intervals on every citation count, acknowledging that the number moves with the same prompt across different days. Profound offers a single blended score with no interval and no run-to-run variance, which is either remarkable stability or remarkable opacity. Gumshoe samples hundreds of queries; Peec samples far fewer. The cost difference is meaningful: Profound costs more than Nightwatch but offers less transparency. This creates a perverse incentive structure: the more expensive your tool, the less likely it is to tell you its methodology. When evaluating a tool, the published interval (if any) is more honest than the point estimate. A tool that says "you have 47% citation share ±8%, across 20 runs" is admitting uncertainty. One that says "you have 51% citation share" is either lying or measuring something different each time without telling you.

Contract terms and hidden costs nobody discloses upfront

Tool vendors publish pricing on their landing pages, but the actual cost of a contract contains terms that are never mentioned until you talk to sales. Understanding these before you talk to a vendor puts you in a much stronger negotiating position.

Minimum contract length is the first hidden variable. Entry-level tools like Peec AI and Rankscale operate month-to-month with no minimums. Mid-tier tools like Gumshoe and Nightwatch typically require three to six month commitments, with annual contracts at a 15 to 20 per cent discount. Premium vendors like Profound require minimum two-year contracts, sometimes with minimum monthly spend floors. If a vendor will not quote you for three months or less, they are not confident you will stick around at renewal—a bad signal about the product.

Overage fees are the second trap. A tool quoting £299 monthly for "unlimited keywords" will fine-print that "unlimited" means up to 500 tracked keywords, and 501 to 1,000 keywords cost an extra 50 per cent per month. A tool saying "daily updates" might mean weekdays only, with weekend data lagging by 24 hours. Read the terms carefully. Many vendors design pricing to make the headline number look cheap, then penalise you when you use features the vendor assumed you would not.

Renewal price escalation is a third hidden cost. A vendor might quote £299 in year one and 115 per cent of year-one renewal price (£344) in year two, regardless of your usage. By year three, that £299/month price has become £395 or more purely through escalation clauses in the contract. On a three-year contract, ask the vendor upfront: is renewal pricing locked to the year-one price, or does it escalate? If it escalates, what is the escalation percentage and can it be capped? A vendor that will not commit to a cap is hiding their true cost.

Refund and cancellation policies rarely appear in published terms. A vendor offering a 30-day free trial might legally have the right to charge you immediately when the trial ends even if you did not explicitly accept a paid subscription (the "negative option" problem). Check whether there is a money-back guarantee if you are unsatisfied within the first 30 days. Most SaaS vendors offer none—payment is non-refundable once the month begins, even if you cancel on day two. This is legal and standard, but knowing it ahead of time changes how you evaluate the product. A tool offering 30-day money-back guarantee is so confident it is useful that it can afford to refund you. A tool offering none is less confident.

Data export and portability are contractual questions most teams forget to ask. If you decide to switch vendors in month six, can you export your historical data? Many tools make data export require a separate data-export fee (50 to 200 per cent of monthly price), or they limit what you can export (summary statistics only, not raw prompts and responses). If you think you might switch vendors, understand the switching cost up front. A vendor that charges £100 to export a year of data is discouraging you from leaving. A vendor that includes free, easy data export is secure enough in their product to assume you will come back.

Seat licenses are often not what they seem. A vendor quoting "£299 per seat per month" might mean five people can view the tool dashboard. But "editing saved searches" might require a higher tier seat at £399/month, and running custom analysis might require a £499/month "analyst" seat. If you need five people to view dashboards, three people to run custom queries, and one person to manage integrations, you are paying 5 × £299 + 3 × £399 + 1 × £499 = £2,916/month, not £299. Ask vendors to spell out what tier each role requires and whether you can mix seat types.

Finally, API pricing and integration costs are almost never included in the headline. A vendor charging £299/month for the tool might charge an additional £100 to £200 monthly (or one-time £500 to £1,000 setup) to enable API access so you can connect their data to your own reporting systems. If you need to integrate their output into Salesforce, Tableau, or Looker, there are often additional fees. Budget for these separately and negotiate them into the contract. A vendor that builds API access into the base plan is saving you money; a vendor that adds it as a line item is extracting more revenue from you than the headline price suggests.

Pricing transparency and what to actually expect to pay

The most critical missing data in the AI visibility market is pricing. Of the 1 tools listed here, 1 do not publish entry-level pricing anywhere on their website. Profound, the billion-dollar market leader, publishes nothing. Peec AI's lowest-tier pricing appears only on aggregator sites like Capterra and G2 reviews, not on their own domain, which means you cannot verify it without leaving Peec's site and trusting a third party. When a category leader hides pricing, it sends a signal: custom sales conversations, minimum annual commitments, and enterprise-only tiers. This is not necessarily bad—it reflects their customer base—but it means a small business or startup will not get a price quote without weeks of sales cycles.

Among trackers that do publish pricing, entry costs range from £17/month (Rankscale) to £300/month (Nightwatch at its starting tier). At the bottom end, you are sampling ten to fifteen keywords monthly and accepting high variance. At the top end, you get daily tracking on hundreds of keywords, confidence intervals, and per-engine breakdowns. The mid-range tools (Otterly at £29/mo, Gumshoe at around £199/mo) occupy the practical sweet spot: real tracking volume without the enterprise price tag. For a company deciding whether to invest at all, the question is not "which tool is best" but "at what price does the tracking volume justify the monthly commitment." If you are tracking ten brand-related queries, Rankscale at £20/mo pencils out. If you are tracking a hundred queries across ten markets and need per-engine visibility, you need Nightwatch or Profound. Most companies fall in between and should ask themselves one question: will this monthly number change our decision-making? If the answer is no, buy the cheapest tracker, run it for three months, and cancel if nothing moves.

Real budget allocation: what it should look like

If you have a £5,000/month budget for AI visibility, allocate it this way: spend no more than £1,000 on tracking tools (once). Spend the remaining £4,000 on the work that actually moves the needle: analyst relations and database inclusion (Chambers, Legal500, equivalent for your sector) at £1,500/month, thought leadership placement (trade publications, syndication) at £1,500/month, and community presence (Reddit, industry forums, Stack Overflow for technical brands) at £1,000/month. Tools will tell you if that spending is working; the spending is what makes you citable. A firm that spends £5,000/month on a tool and nothing on the underlying work will see its score improve temporarily, then stagnate, because it has made itself no more worth citing than before. A firm that invests £1,000 in a baseline measurement and £4,000 in actual visibility work will see change that compounds, because it has changed what AI systems have available to cite.

Total cost of ownership: what you will actually spend

A tool's headline price is never the full cost. Peec AI's entry tier at £99/month looks cheap until you factor in setup (reading API documentation, configuring integrations, designing your monitoring dashboard) at roughly 8 hours of time, and ongoing interpretation (weekly reviews, manual audits to spot when the tool is lying, chasing down discrepancies between tools). A typical mid-market firm spends three to five hours weekly on this work. At £40 per hour (loaded cost), that is £120 to £200 monthly in labour cost alone, doubling the apparent software spend. Nightwatch at £299/month appears expensive until you account for the fact that it publishes confidence intervals and supports custom query sets, reducing the audit overhead to one hour weekly. The blended cost becomes £299 + £40 = £339, versus Peec at £99 + £160 = £259. But Nightwatch's confidence intervals mean your stakeholders stop asking "why did we drop 2 points?" questions that consume another two hours of investigation monthly. By month three, the labour savings have flipped: Nightwatch costs less in total time spend because it communicates uncertainty openly.

Cost per tracked metric: the real pricing breakdown

Comparing headline pricing ignores what you actually get for the money. When evaluated per tracked keyword and update frequency, tool costs diverge sharply. Rankscale at £17/month tracks roughly 10 to 15 keywords weekly; that is £1.13 per keyword per month for weekly data. Peec AI at £99/month with their standard tier covers about 20 to 50 keywords depending on customisation; call it £2 per keyword monthly for weekly tracking. Gumshoe at £349 month covers 200+ queries with daily updates; that works to £1.75 per keyword per month but with daily granularity, not weekly. Nightwatch starting at £299 covers 100 tracked keywords with daily updates and confidence intervals; £2.99 per keyword monthly but with methodology transparency. For an enterprise on Profound's unlisted pricing (typically £500 to £2,000 monthly based on customer interviews), tracking 500+ queries with daily data works to approximately £1 to £4 per keyword monthly depending on contract scale. The cheaper tools win on per-keyword cost only if you track fewer than fifty queries; the premium tools win on transparency and integration if you track hundreds. A firm that can articulate exactly which queries matter (typically 20 to 50) can use the cheapest tool profitably. A firm that wants exhaustive coverage across a market cannot.

Tool pricing tiers and what each delivers

Different budget sizes support different tool choices. For a bootstrapped startup or freelancer with under £200/month to allocate to AI measurement, Peec AI at £99 to £199 depending on tier makes sense—you get basic brand mention tracking across a limited set of prompts. You sacrifice frequency (rescans happen weekly rather than daily) and sample size (fewer queries), but the cost aligns with the idea that you are measuring a baseline, not obsessing over month-on-month movement. Peec's approach is honest about this limitation: they make no claims of hourly precision.

Mid-market firms with £300 to £500/month in the budget should compare Gumshoe (£349) and Nightwatch. Both test hundreds of queries, both publish their methodology, and both can show whether you are actually cited in the major engines or just in the noise. Gumshoe skews toward high volume; Nightwatch publishes confidence intervals on every figure, acknowledging that repeated runs return different results. For a firm that needs to explain variance to stakeholders ("why did our score drop 3 points when we did nothing?"), Nightwatch's transparency is worth the cost.

Enterprise buyers (£1,000+/month) are shopping for integration with existing tools, custom query sets, API access for internal dashboards, and dedicated support. Profound at the high end and Nightwatch's enterprise tier are the only choices at this scale that have actual feature differentiation, though Profound's opacity on pricing and methodology remains a red flag even for large buyers. If your team has an engineering resource to build a custom dashboard, a tool's API access matters more than its dashboard UI.

The decision should not be "what is the best tool?" but "what do I need to measure, and does this tool's cost justify its answer frequency and sample size?" A tool delivering daily updates across 300 queries is wasted on a brand that publishes once per week. A tool delivering weekly updates across 20 queries is insufficient for a brand testing hypothesis ("if we get into TechCrunch, does Perplexity citation rise?") because the noise floor is too high to detect change.

Every claim in this section is graded and sourced in the evidence ledger. Our commercial interest in each tool named here is published on who pays us.

Pricing tier decision matrix: when the premium cost actually justifies itself

The jump from entry-level (£99-299/month) to premium (£1,000+/month) tools often fails the ROI test for organisations without clear use cases. Here is when the upgrade makes sense. First: if you are testing whether a specific tactic works ("does analyst inclusion move Perplexity citations?"), you need a premium tool with enough query coverage (200+ tracked prompts) and refresh frequency (daily or near-daily) to detect signal above noise. Entry-level tools with weekly rescans on 20 keywords cannot distinguish a real effect from random volatility. If you are hypothesis-testing, premium tools pay for themselves through faster validation. If you are reporting to stakeholders monthly without a hypothesis ("here is our AI visibility score"), the premium cost is pure overhead. Second: if your brand operates across multiple markets or product lines, each with different target audiences and AI search patterns, you need tool sophistication enough to track separately. A SaaS company selling to SMBs in the UK and enterprise in North America needs visibility tracking per-market, which requires 100+ custom queries and per-engine breakdowns. Enterprise tools provide this; entry-level trackers do not, and trying to synthesise results across multiple entry-level tool instances becomes expensive in labour. A single-market or single-product brand, by contrast, gets diminishing returns from premium tracking. Third: if your measurement infrastructure already feeds other systems (your BI platform, your marketing stack, your board reporting), a premium tool's API access justifies integration cost in reduced manual reporting. Entry-level tools give dashboards; premium tools give data. The difference matters only if you already have infrastructure to absorb that data. A firm without a BI team, trying to integrate tool output into existing analytics, will find the premium tool's API worthless. For these three cases—no clear hypothesis, single product or market, no existing BI infrastructure—stick with entry-level tooling. The premium tier will not change your decision-making; it will just increase your costs and create the perception of sophistication where none is justified.

Pricing by business type: which tool makes financial sense where

Choosing a tool is not about which is "best" in the abstract; it is about which solves your specific measurement problem at a price you can justify. A freelancer or micro-agency with one to five clients should not buy the same tool as an enterprise managing AI visibility across fifty brands. Freelancers and solopreneurs (£0-500 monthly AI measurement budget) have only one viable path: do not buy a tracker at all. Instead, run manual checks on ChatGPT, Perplexity and Google AI Overviews yourself once per week for the five to ten queries that matter to each client. Spend zero and keep the data in a Google Sheet. This works because you have low volume and can remember the results. Your time is cheaper than a tool subscription. Micro-agencies serving five to twenty clients (£500-1,500 monthly budget) should choose between Rankscale at £17/month or Peec AI at £99-£199. Rankscale covers up to 20 tracked keywords at weekly rescans; Peec covers more keywords with flexible query selection. Either fits in a £100 monthly allocation and scales linearly: more clients means more tool seats, not exponential cost. Small businesses and startups (£1,500-5,000 budget) operate in a different calculus. You need to measure whether your own efforts (publication placement, Reddit participation, analyst inclusion) are actually moving citations. Gumshoe at £349 or Nightwatch at £299 enters this range. Both offer daily tracking on hundreds of queries, which lets you spot whether a PR campaign or publication placement moved your visibility. The investment justifies itself if you run a A-B test: "if we get into TechCrunch, do citations rise?" or "if we launch on Product Hunt, do Perplexity mentions increase?" At this scale, the tool budget is part of a hypothesis-testing infrastructure, not a vanity metric. Mid-market companies (£5,000-20,000 budget) have enough volume to justify premium-tier spending. Nightwatch's enterprise tier or Gumshoe's mid-tier serve companies tracking 300-500 keywords across multiple markets. The annual contract (£3,000-6,000) is defensible if you have multi-regional operations or track visibility across product lines. Enterprise buyers (£20,000+ budget) should demand custom integration, historical data access, and API connectivity. At this scale, Profound becomes an option, though its unlisted pricing typically runs £1,000-5,000 monthly depending on query volume and contract length. The tell: if your company is big enough to have a Searcher report upward, you are big enough to allocate dedicated tool budget. Below that threshold, do not.

TCO reality: when the cheapest tool is actually the most expensive

A tool costing £99 monthly appears cheaper than one at £399 until you account for what each actually delivers. Peec AI at £99 requires you to interpret a dashboard yourself; if you spend 8 hours monthly trying to understand whether a 5-point citation movement means anything (spoiler: it does not), that is £320 in labour cost (8 hours at £40/hour), making total cost £419 monthly. A Nightwatch subscription at £299 publishes confidence intervals automatically—"your citation share is 42% ± 7%, across 20 runs"—which answers the interpretation question for you and saves 6 to 8 hours of investigation monthly. Total cost becomes £299 + £80 (one hour of review time) = £379, cheaper than Peec despite the higher headline price. The worst TCO trap: a £17/month Rankscale tracker for someone tracking 500 queries. The tool samples only 10-15 keywords because of its pricing model, so you add a second tool (another £17), then a third, then switch to Gumshoe, then regret not starting with Gumshoe. Total spend over a year: £600 in platform switching costs plus the psychological cost of incomplete data. A firm that starts with the right tier for its volume saves money and sanity. When evaluating a tool, ask: "How many hours per week will we spend interpreting this data?" If the answer is more than two hours, the cheaper tool is actually expensive in labour. If the answer is zero because the tool shows you clear signal/noise boundaries, the expensive tool is cheap.