Reference

Answer engine optimization, measured

We sell a competing product. We are telling you to start with theirs because at the entry tier the arithmetic says so, and because a comparison page that always concludes “us” is worth nothing to you.

What does answer engine optimization actually cost per answer?

Every AEO tool does roughly the same thing underneath: it asks an engine a question, reads the answer, and looks for your name. The cost of that unit is public. Anthropic charges $10 per 1,000 searches for the web search tool, plus token costs. Google charges $14 per 1,000 grounded queries on Gemini 3 models after a monthly free allowance, and bills each search the model decides to run, so one prompt can bill several times.

So one grounded answer costs somewhere near a cent to a cent and a half in search fees alone, before a single token is paid for. Hold that number. It is the floor under every price on this page.

Why price per prompt is not a comparison

Vendors sell prompts. A prompt is not an answer. A prompt tracked daily across four engines is 120 answers a month. The same prompt tracked weekly across one engine is four. Two products can advertise 100 prompts and differ by a factor of thirty in what they actually buy you.

Run the division and the spread is large. Otterly’s Lite plan is $29 a month for 15 prompts across four engines with daily tracking, about 1,800 answers, or roughly 1.6 cents each. Profound’s Starter at $99 a month billed annually tracks 50 prompts on ChatGPT only, about 1,500 answers, or roughly 6.6 cents each.

Profound is not four times worse value than Otterly. It is selling something different: a consumer panel and a Conversation Explorer built on real query volume, from a company that raised a $96M Series C at a $1B valuation in February 2026. The per-answer number tells you what monitoring costs. It does not tell you what a panel is worth.

The one vendor that prices in answers

Rankscale is the only product in this set that publishes its price in the unit that matters. Pro is $99 a month for up to 4,800 tracked answers, Growth $385 for up to 22,000, Enterprise $780 for up to 48,000. That is 2.1, 1.8 and 1.6 cents per answer, landing almost exactly on the list cost of the underlying grounded call. That is what a defensible price looks like, and it is worth noticing that the vendor who made the unit legible is also the one whose margin is easiest to check.

We could not make Evertune’s arithmetic resolve. Their Pro plan is $800 a month for 100,000 prompts across 11 models, sampled up to 100 times per model. Taken literally that is over a hundred million answers for $800, which cannot be what the words mean. We are flagging it as unresolved rather than guessing, and it is a fair question to put to them on a call.

Which AEO tool is the best value for money right now?

Ranked on cost to get a defensible reading of your own named prompts, for a company spending under $300 a month. All prices read from the vendor’s own page on 2026-08-20.

Two products sit outside the ranking because they are not subscriptions. HubSpot’s AI Search Grader is a free one-time check. It costs nothing and takes two minutes. Run it before you buy anything. And Peec AI is widely quoted at around $95 a month, but their pricing page did not render a price to us on 2026-08-20, so we are not carrying a number we could not read ourselves.

When is the cheap tool the wrong choice?

When you need to explain a change to someone who will ask how sure you are. Rand Fishkin’s team ran 12 prompts across three engines with 600 volunteers, 2,961 runs in total. The result: there is less than a 1 in 100 chance that two responses to the same prompt return the same list of brands, and about 1 in 1,000 that they return the same list in the same order.

Fishkin’s own conclusion is the part most people skip: visibility measured as a percentage across dozens or hundreds of prompts, run multiple times, is a reasonable metric. The metric is fine. The single run is not. A 2026 paper by Schulte, Bleeker and Kaufmann makes the same argument formally and asks the field to treat visibility as a distribution rather than a point.

This is where the $29 plan is quietly stronger than it looks and quietly weaker than it looks. Daily tracking gives you thirty samples a month for free, so your monthly average is defensible. What it cannot do is tell you whether Tuesday’s twelve-point drop was real, because Tuesday is n=1. If your reporting cadence is monthly, cheap and daily is fine. If it is weekly, you are reading noise.

What should you check before you pay for any of them?

How many runs is each number built on? Almost nobody prints this. Ask for it in writing before the trial ends. Can you see the raw answer? A percentage with no answer behind it cannot be audited or shown to a sceptical executive. Who chose the prompts, and who chose the competitors? If you wrote both lists, you are measuring a pool you built.

Is the vendor’s own advice tested? Most of the standard AEO checklist has never been measured against a control. Ahrefs tracked 1,885 pages that added JSON-LD against roughly 4,000 matched controls and found AI Overview citations moved minus 4.6%. Google has separately said it does not use llms.txt and does not plan to.

What does appear to work is unglamorous. Listicles account for 63% of AI citations across roughly 25,000 of the most-cited URLs, and 71% to 86% of those listicles are ranked rather than unordered. Note that Search Engine Land labels that piece sponsored vendor content. In a separate study, the typical heavily cited page runs about 941 words with 4 H2s, 2 H3s, 15 external links and 10 images.

Where does Rovoki sit in this, honestly?

Rovoki measures four engines, ChatGPT, Perplexity, Gemini and Claude, reported separately and never averaged without the per-engine split beside it, across 18 countries and 31 languages or worldwide with no country pinned. It returns a Visibility Standing from 0 to 100 and publishes the raw answer behind every question it counted.

What it does not do yet: it samples n=1 per run and does not publish a confidence band. Bands are what we are building toward and they are not shipped. If sampling depth is your buying criterion today, Evertune samples each prompt up to 100 times per model and you should buy that instead. If your problem is that you cannot show anyone the evidence behind the number on your dashboard, that is the part we built first.

The category is filling up fast, which is good for you. G2’s answer engine optimization category went from 7 products in March 2025 to more than 150 within ten months. Prices are falling. Wait a quarter and this page will need rewriting, which is why it carries a date.