What SEO data actually costs, measured call by call

The short answer

Here is what each call behind an SEO tool costs, in August 2026. Where a row says measured, we diffed our own provider balance before and after a real call. Where it says list price, it is the provider's published rate and we have not tried to verify their invoice against it.

Nobody publishes these next to each other, so the shape is easy to miss. A rank check and an audit page cost fractions of a cent. A keyword volume lookup costs about 190 times a standard-queue rank check, and the old way of measuring AI visibility cost about 170 times one.

Two of those numbers changed how our product works. Both are below.

How we measured

Provider price lists are organised by product, not by the thing you actually do. So we did the boring version: record the account balance, make one real call, record it again, subtract.

That method is why we can say $0.00045 per audit page rather than "about $0.0006", which is what we had modelled before anyone checked. It is also how we found the number that was wrong in shape rather than in size.

The caveat is on the record too. A six-page crawl cannot separate a fixed per-crawl fee from a per-page fee, so $0.00045 is an upper bound, and we use it that way in the budget that guards our free tier.

Keyword volume is priced per call, not per keyword

Every cost model we had assumed keyword enrichment was priced per keyword, at about $0.0005 each. It isn't.

DataForSEO's own documentation for the Google Ads search volume endpoint says it plainly:

Our system will charge your account per each request, no matter what number of keywords an array has, the price for 1 or 1000 keywords will be the same.

We confirmed it on the balance. Five keywords: $0.09. Twenty keywords: $0.09. Adding the two Labs endpoints we call alongside it brings a full enrichment to about $0.115 per call.

So the same ten keywords cost ten times more if you add them one by one:

That gap explains a lot of otherwise strange behaviour in keyword tools you already use:

• Bulk paste boxes that ask for a list rather than one keyword at a time. The box is a cost control wearing a convenience feature's clothes.
• Volume data that refreshes monthly on a schedule instead of when you ask. One scheduled batch for the whole account is one call.
• Free tiers that show volume for a fixed starter list but never for a keyword you type. The typed keyword is a fresh call at full price.
• Tools that quietly cap "keywords per lookup" high rather than low. A cap of 1,000 costs the vendor exactly what a cap of 5 costs.

If your tool bills you per keyword and buys per call, the batching is where its margin lives.

Why our own free tier has no keyword volume at all

We publish our free tier's budget: a free account may cost us $1 in provider spend for the lifetime of the account, enforced in code and asserted at build time.

Against that ceiling, a small free keyword allowance was impossible. Enrichment costs $0.115 per call whether the allowance is one lookup or a thousand, and a usable allowance came out at $1.38 of a $1 budget. So free accounts get zero keyword volume, difficulty and CPC, not a small amount. Free tracks keywords and checks rank weekly. Volume data is what a paid plan is for.

The reason is arithmetic, not stinginess. A free tier that spends its whole budget on one feature is a free tier that gets withdrawn three months later, and we would rather ship one that survives.

AI visibility got about 67 times cheaper in one provider change

We used to measure AI visibility through DataForSEO's LLM mentions endpoint, at $0.10 to $0.116 per call. That endpoint scans a pre-collected index of AI answers. You are reading someone's earlier crawl.

We now ask Google the question live through Bright Data's SERP API, which lists $1.50 per 1,000 requests with 5,000 requests a month free. That is $0.0015 per question.

The cheaper option is also the better data. Bright Data's AI Overview API returns the overview text with the sources it cites, so we get the actual citation URLs and can tell you which domains Google named instead of you. That is the whole question worth asking, and the expensive endpoint could not answer it.

A nine-question scan of stripe.com cost us $0.0135. That is the entire provider cost of finding out who Google's AI names in a category, and the method behind it is one you can run yourself.

What this implies about the tools you buy

Take the Google AI Overview line on its own, because it's the only one we can price from public numbers, and run the arithmetic on published plans.

A plan tracking 25 prompts daily makes about 750 checks a month. At $0.0015 a check, that is $1.13 of provider spend. A plan tracking 15 prompts daily is about 450 checks, or $0.68.

Two honest qualifications before anyone draws the wrong conclusion. First, tools that also check ChatGPT, Perplexity and Copilot are paying for LLM API calls we cannot price from a published rate card, and those are more expensive per answer than a SERP fetch. Second, low data cost is not an accusation. Salaries, storage, alerting and support are most of what any software company spends, and pricing on data cost would be a strange way to run a business.

What the arithmetic does buy you is a better question at a sales call. Not "why does this cost $500", but "how many checks a month is that, on which engines, and how often". A vendor who can answer that in one sentence is selling you a product. One who cannot is selling you a score.

What our own tiers cost us

Since the point of this post is that nobody publishes this, here is ours, computed from tier limits rather than a spreadsheet:

The build fails if any paid tier drops under a 50 percent margin floor, which is why Agency is $229 and not $199. At 50 tracked questions and 1,200 tracked keywords, $199 did not clear the floor.

Corrected 28 August 2026. This table used to read $19 / $49 / $179 at margins of 83, 65 and 70 percent, and every one of those margins was wrong in the direction that flattered us. Our rank-check rate was modelled at $0.0006 from DataForSEO's published "$0.60 per 1,000", which prices one page of ten results. We ask for the top 100, so we were buying ten pages and costing one. Measured off the provider's own cost field on 28 August: $0.00465 at depth 100 and $0.00105 at depth 20. At the real rate the old Pro was running at minus 66 percent. The prices above are what it takes to clear the floor once the rate is right, and the tiers now check the top 20 routinely with a full top-100 pass on Agency, because depth is what the bill is actually for.

Lifetime is the row that shows why this matters. $499 covers 37 months of provider spend, priced against the AI rate that takes effect in January rather than today's, because a one-time payment cannot be repriced later. Every future increase to the AI visibility allowance is free for those accounts forever, which is why it's capped at 10 seats rather than sold as a launch promotion.

What to do with this

1. Add keywords in batches, not one at a time, in whatever tool you use. If it bills you per lookup you will feel the difference immediately, and if it doesn't, your vendor still does.
2. Before you pay for AI visibility tracking, count the checks. Prompts times frequency is the number that decides value, not the plan name. The comparison of AI visibility tools has that arithmetic for six products.
3. Ask any vendor what one scan costs them. The answer is not confidential, it's on their provider's price list, and the reaction to the question tells you plenty.
4. If you want to run the free checks yourself first, the free tools cost nothing because none of them calls a paid provider.

Rates in this post were measured or read from published price lists in August 2026, and they move. The AI visibility rate moved by a factor of 67 in a single provider change, which is the best argument we have for measuring rather than modelling.