Our 41% Was Wrong. We Counted Seven Tools That Publish Both Numbers, and Eighteen Did.

This article opened with a number that was wrong. It said the entry plan gives you 41% of the engines a vendor advertises. The real figure is 65%, and the mistake was ours, not the market’s.

Here is what happened, because the mistake is more useful than the number.

We counted seven tools that publish both. At the moment we published, eighteen tools in our own catalogue carried both an entry-plan engine count and a top-tier one, and ten of those eighteen had no gap at all: the same coverage on every tier. Our sample contained the seven that made you climb and none of the ten that did not.

That was not a judgement call, it was a data-structure accident, and it is the kind worth describing in detail (see how we got it wrong).

Recounted properly today, across the 53 tools that publish both figures: entry plans carry 211 engines between them and top tiers carry 335. That is 63%, not 41%.

How this was compiled: every figure comes from our verified tool catalogue, where each entry records the vendor’s own pricing page URL and the date it was read. Nothing here is a survey, an estimate or a vendor claim about a rival. Where a vendor publishes no number, the tool is counted as publishing no number rather than guessed at. Recounted on 8 August 2026 against 80 purchasable tools, and the price table recounted again on 10 August 2026 after it was found to be three rows short of its own stated rule.

The short version

1. 65% of advertised coverage on the plan that has a price, across the 53 tools that publish both figures. We previously said 41% from a sample of seven.
2. Twenty-four of those 53 never raise the count at all. Whatever you pay, you get what the entry plan gave you. We previously said two.
3. Eleven of the 80 publish no entry price at all, and five publish no engine count either.
4. Claude is on the entry plan at eight vendors and costs extra or is unavailable at four. We said the opposite and we were wrong about that too.
5. Price is a weak predictor of coverage, not a null one. Across the 53 tools that publish both figures the rank correlation is +0.28. The eleven rows we first printed gave +0.05, and that is the number this article was built on.

The gap between the headline and the plan

Fifty-three tools publish both an entry-tier engine count and a top-tier one. 27 of them make you climb. One of them climbs in a shape this table cannot show: Limy sells a number of engines rather than a list, so it reaches five at the top and three at entry like any other row, while the plan in between costs three hundred dollars a month more than entry and grants the same three, chosen from a pool of five. 6 other tools let you pick your engines from a pool, so the packaging is not unusual; the price of the pick is, at three hundred dollars a month for the same three engines.

The plainest instance of the pattern this article is about arrived last: GeoGenie sells two engines on its $100 plan, three on the $200 one, three again on the $500 one, and all five only on the tier that publishes no price. The $300 step in the middle buys prompts and brands and not one additional engine, which is the shape of this table stated as a price list.

The 27 that do make you climb:

The other twenty-six do not raise the count at any price: Alhena, Athena HQ, DinoRANK, Finseo, Geoptie, Gumshoe, HubSpot AEO, LLMrefs, Meteoria, Nightwatch, Otterly, Qwairy, Radarkit, Rank Prompt, Rankscale, ScalePost, SE Ranking, Semrush, seoClarity ArcAI, SERPs, Serpstat, Similarweb, SISTRIX, Superlines, Waikay and ZipTie.

Add all 53 together and entry plans carry 211 engines against 335 on the top tiers. That is 63% of the advertised list, and a 37% gap. The 65% an earlier version printed here never matched its own two numbers, and this recount is where it was caught.

Those two totals were 175 and 257, then 205 and 329, and vendors added since have moved them again. Scrunch carried all seven engines it then supported on every paid tier, which put it in the no-gap list above. Between 7 and 11 August 2026 it withdrew its two self-serve tiers, replaced them with a single $250 plan, added Copilot and Grok to the product, and put five of the nine engines behind a custom contract. So its top-tier count rose by two, its entry count fell by three, and it crossed from one side of this article’s central split to the other in four days. The review has the before and after side by side.

That is worth stating plainly because it changes how much weight this whole comparison can carry. The split between vendors that gate engines and vendors that do not is not a property of a vendor, it is a property of a price list on a date. Every figure in this article is dated for that reason, and the two totals above are recomputed from our catalogue on every build rather than typed, which is the only reason this update happened at all: the build refused with “the catalogue now sums to 172” rather than letting the old pair sit here looking correct.

Two of the climbers deserve reading twice. Surfer’s Standard plan at €99 tracks one engine where the Pro tier tracks five. Writesonic’s $79 plan tracks three of the ten platforms it advertises, and the other seven need Enterprise, which publishes no price at all: so the difference between the headline number and the buyable number is unpriceable by design.

And Dageno does something stranger. Its Starter, Growth and Scale plans all track three engines out of seven. The count never rises with the price. Only Enterprise says “full coverage”, and Enterprise is unpriced. You can pay several times more and receive exactly the same coverage.

How we got it wrong

The seven tools that the original version counted were not chosen. They were what a field returned.

Our catalogue records a vendor’s top-tier coverage in a field called topEngines, and that field’s own documentation said it holds the count “when it reaches more than the entry plan”. Written for a reason: there is no point storing a second number identical to the first. The consequence is that the field was only ever populated for vendors with a gap.

Then we asked the catalogue which tools publish both counts, and it answered with the climbers, because for everyone else the second number had never been written down. We measured the size of a gap using a sample selected for having one. The ten flat vendors were sitting in the same file with their coverage recorded once instead of twice, and were invisible to the question.

Nothing was fabricated and no vendor was misrepresented. Every individual review was right. The market statistic built on top of them was wrong by 27 points, in the direction that made the market look worse than it is.

Two things changed as a result. The catalogue now records a top-tier count for every vendor that publishes one, whether or not it differs from the entry count, so the sample is the population. And this article carries the recount rather than a correction buried at the bottom, because the original number is the one people quoted back to us.

The claim we got wrong

In our Nightwatch review we wrote that Claude on the entry plan is unusual, “which most of this market puts behind Enterprise”. We had read a handful of pricing pages and generalised.

Counting all 39 pricing pages we had read on 7 August 2026 instead, which is where Claude sat that day:

Seven vendors put Claude on the plan you can buy. Three put it behind Enterprise, and one names it on its homepage while selling it on no tier at all. The claim was the wrong way round, it has been corrected in that review, and an impression formed from four pricing pages did not survive thirty-nine.

Two details that survived the recount. Rank Prompt inverts the usual line entirely, putting Claude inside its base credit while charging extra for Gemini. And Nightwatch’s real distinction is not that it includes Claude but that its engine list never changes as you climb. That looked rare when we wrote it. It is not: twenty-six vendors in the catalogue never raise the count, Nightwatch among them, which is the second thing this recount overturned.

Price barely buys coverage, and the table we printed said it bought none

Sort the catalogue by what the entry plan actually tracks and the price column almost stops making sense. Almost is the correction, and it is the second one this article has had to make about its own sampling. It is now the third, and the third one is described in full below rather than absorbed quietly.

The rule for this table: every tool in the catalogue with both a published entry price and a stated engine count on that plan. Fifty-five qualify. Nothing is left out for being inconvenient, prices are printed as each vendor prints them, in the currency it charges in, and none are converted.

Rank those fifty-five by price and by engines and the two orders agree, weakly: Spearman’s rho is +0.28. Paying more does buy a little more coverage, on average, and the honest summary is that it is a weak predictor rather than no predictor at all.

The first version of this section printed eleven of those rows and concluded that price buys no coverage whatsoever. Those eleven give rho +0.05. We can date the error precisely, because the catalogue is in version control: when this article was published at 11:16 on 7 August 2026 the catalogue held thirty-two tools, twenty-five of them qualified for this table, and those twenty-five gave rho +0.53. The same eleven gave +0.05 then too. Our own data contradicted the claim on the morning we made it, and by a wider margin than it does now.

There is no stated rule for why those eleven and not the other fourteen, because there was not one. That is the same failure as the 41%, in the same article, found the same day: a number computed over whatever subset was in front of us and reported as if it described the market.

And then this article did it a third time, to itself. On 10 August 2026 we recounted the table above against the catalogue and found it printed thirty-eight rows while the sentence introducing it claimed thirty-seven and the rule stated in that same sentence returned forty-one. Three numbers, one table, none of them agreeing. AIclicks, LLMrefs and Search Atlas all publish an entry price and an entry engine count, all three met the printed rule, and all three were missing. They are in the table now.

The correction moves rho from the +0.32 we printed to +0.28, which does not change the conclusion and is not the point. The point is that a stated rule and a hand-maintained table drift apart silently, and the only thing that catches it is recounting the table against the data rather than rereading the prose. This article has now been wrong about its own denominator three times: once by sampling a field that only existed when there was a gap, once by printing eleven rows of twenty-five with no rule at all, and once by letting a table fall three rows behind the rule written directly above it. The first two we found because a reader would have. This one we found because a guard now counts the rows, and that guard exists because of the first two.

The currency mixing does not rescue it either. Treating every euro price as up to 1.20 dollars moves rho from +0.284 to +0.272, and leaves the eleven-row figure unchanged at +0.05.

What does survive is the shape at the extremes. The cheapest tool you can pay for, at $20, tracks ten engines. The one at $499, twenty-five times the price, tracks five. The widest coverage you can buy at all is ten engines, sold by Rankscale at $20, by LLMrefs at $79 and by Qwairy at €79, and it beats a $2,500 platform’s nine and a $295 plan’s eight. At $79 or less the median entry plan carries three engines; above $79 it carries five. That gap of two engines is the entire size of the effect.

We are not saying the expensive tools are bad value. They sell depth, seats, integrations, managed services and support that a $29 tracker does not. We are saying engine coverage is a weak reason to climb a pricing page, worth about two engines across the whole range, and that the tiers are arranged to look like it is a strong one.

What eleven vendors will not tell you at all

Eleven of the 80 publish no entry price: AirOps, Peec AI, Evertune, Conductor, Cloudflare AEO, Quattr, ScalePost, XFunnel, Brandlight, Bluefish AI and BrightEdge.

Five publish no engine count either: Evertune, Brandlight, Bluefish AI, Conductor and Cloudflare AEO. Neither number, on either axis a buyer would compare. A sixth, Rankability, publishes four prices and no engine count at all, which is the rarer failure: it tells you exactly what to pay and never what you are measuring on.

Add the one whose AI product is unpriced on top of a priced platform, seoClarity ArcAI, and roughly a fifth of this market cannot be compared without a sales call.

There used to be a pattern inside that, and it has weakened: the widest engine lists and the missing prices travelled together. ArcAI publishes nine engines and a $2,500 platform floor, XFunnel nine and no price. But the highest published entry count you can buy outright is now ten, sold by Rankscale at $20, by LLMrefs at $79 and by Qwairy at €79, which breaks the rule about as cleanly as it can be broken: the widest entry list in the catalogue is also its cheapest plan, and the two vendors who match it charge under $80.

What to do with this before you buy

Ask which engines are on the plan you are quoting, not on the product. The gap averages 35% of the advertised list across the tools that publish both, and it is 0% at twenty-four of the 44.

Ask whether the count rises with the price. At twenty-four of the 53 vendors that publish both numbers it never rises, which is worth knowing in both directions: you are not being upsold, and you cannot buy your way to more.

Decide which engines you actually need first. Google’s own surfaces cite almost nothing in common, so “Google” is not one item, and a tool covering nine engines badly is worth less than one covering the three your buyers use.

Then ask how many times each prompt is checked. Coverage is one axis and repetition is the other: the number of runs a category needs runs from about three to about thirty, and a wide, shallow tool answers the wrong question confidently.

What this does not show

• It is a census of published claims, not of products. Every figure is what a vendor printed on a given date. A tool may cover more than it says, or less.
• 80 purchasable tools is our catalogue, not the market. We add vendors when they have measurable demand or turn up in the answers we measure, and we say so.
• The first version of this article got its headline number wrong, by 27 points, from a sample of seven where eighteen were available. The method that produced it is described above rather than deleted.
• Prices move. Each entry carries the date it was read; several were re-read this week and changed.
• Engine counts are not comparable across vendors. One counts “Google” as one item, another names AI Overviews and AI Mode separately, and a third names a company where the product is an assistant. We record what each page says.
• Nothing here measures quality. A tool that tracks nine engines badly and one that tracks four well look identical in these tables.

Common Questions About AI Search Monitoring Pricing

How much do AI search monitoring tools cost?

Published entry prices in our catalogue run from free to $2,500 a month, and 59 tools publish an entry price under $200. Eleven publish no price at all. Every figure carries the date it was read from the vendor’s page.

Do more expensive tools track more AI engines?

Not reliably, and we got this wrong once already. Across the 55 tools that publish both an entry price and an entry engine count, the two rank-correlate at +0.28, which is weak. The cheapest tool you can pay for tracks ten engines on a $20 plan, while a $499 plan tracks five of the six it names and a €119 plan tracks four. At $79 or less the median entry plan carries three engines, and above $79 it carries five. Two engines across the whole price range is the entire effect.

Which AI engines are usually missing from the entry plan?

There is no single answer, which is the finding. Across the 53 tools that publish both counts, entry plans carry 211 of the 335 engines their top tiers carry, or 63%. An earlier version of this article said 41%, from a sample of seven that excluded every vendor with no gap. Claude is on the entry plan at eight vendors and costs extra or is unavailable at four, so it is not the reliable upsell we previously assumed. Google’s AI Mode and Copilot are the ones most often absent from smaller tools entirely.

Why do some tools not publish a price?

We can report the pattern, not the reason. Eleven of the 80 publish no entry price, five publish no engine count either, and one more sells an unpriced AI module on top of a priced platform. Those tools are aimed at enterprises, where contracts are negotiated, and two of them carry the widest published engine lists in the catalogue.

How should I compare two tools on coverage?

Get the engine list for the specific plan you are being quoted, in writing, with the products named rather than the companies. Then check whether that list changes on higher tiers, because at twenty-four of the 53 vendors that publish both numbers it does not change at all. Finally check the refresh frequency, since coverage without repetition tells you very little.

Where does this data come from?

Our verified tool catalogue, which records each vendor’s pricing page URL and the date each figure was read. It exists because an earlier version of that comparison stated prices as prose and went five months without anyone checking them, by which point it recommended a tool that had shut down.