The price shows before you pay

Technical information first: a custom report on AEOSearch is priced one engine at a time, and the page shows every line before anyone types a card number. Switch on ChatGPT, Claude, Gemini, Perplexity or Grok, pick the model version each one answers on, choose 20, 50 or 100 buyer questions, and the rail on the right rewrites itself on every click: one line per engine with the model named beside it, a $2 line for the questions and the diagnosis, and a total. Nothing on that rail is an estimate. The server hands the page the same per-question rates it uses itself, and when you pay it re-derives the price from its own catalogue and refuses any checkout whose amount disagrees, so the figure you clicked on is the figure Stripe is told, to the dollar.

We didn't have to do it this way, and most of the category doesn't. The usual shape is a tier, a seat count and a sales call, with the arithmetic kept on the vendor's side of the table. We've run thousands of grounded answers through five engines and read every bill line by line, so we know what a question costs on each of them, and we decided that knowledge belonged on the page rather than in a spreadsheet only we could see. What a grounded answer costs is already published; this post is the decision that followed from it.

A price is a claim, and claims need receipts

Every number in one of our reports traces to a stored answer, with the model that gave it named and the sample count printed on the report: that's the whole product, and it would be strange to hold the price to a lower standard than the score. So the price has the same anatomy the score does. Each engine's line is the per-question rate for the model you chose, times the questions you chose, and the rate covers the three passes we take of every question unless you choose the single pass, plus the extraction pass that reads each stored answer, so there's no second line that appears later. Change the model and the line changes: GPT-5.6 Sol and Claude Sonnet 5 read a great deal of retrieved context per answer and cost accordingly, Gemini 3.6 Flash and Perplexity's Sonar cost a fraction of that, and the page shows the difference instead of averaging it into a bundle.

The frontier three at 20 questions come to $180, the same report the fixed ladder sells at $180, and the 50 and 100 buttons show their own prices for that configuration, $445 and $885, before you press either. All five engines at their frontier rungs on 20 questions is $250, which is what Full Coverage costs on the same page, because a spec shouldn't cost two different amounts depending on which door you came in through.

An opinion we'd defend in a room full of people who sell software: hidden pricing is a tax on the buyer's time, and a buyer who has to book a call to learn what a report costs is being asked to pay twice. We'd rather lose the deal we might have won by obscuring the number than win it by hiding it, and we'd put money on the buyers we want agreeing with us.

What it cost us

An admission, because a price this visible cuts both ways. Publishing the per-model rate means publishing the shape of our margin: anyone who has read our cost post can set the two side by side and see roughly what we keep, and there's no way to show one without implying the other. We decided that was fine. What we keep pays for the question generation, the diagnosis, the storage of every answer verbatim, the settlement that refunds failed calls, and the people who read the reports, and a buyer who can see the arithmetic can also see that none of it is padding.

It also means we can't quietly change a price. A rate on the page is a rate a buyer can screenshot, so every move in the catalogue is a move in public, and a model our adapter has never answered on is shown on the page as not yet available rather than sold anyway; a rung we've run but not yet probed for cost is priced from its vendor's sheet with a margin and labelled list price, so the page says which kind of number each line is. That last rule costs us the odd sale: Perplexity's Sonar Pro and its reasoning rungs sit visible and unbuyable today, marked "not yet available", because a report we take money for may not depend on a model we've never called.

And the rounding is on the page too, so we'll say it here as well: a quoted price rounds up to the next $5, never down, and no report sells under $25. $87 becomes $90. We'd rather you read that from us than notice it on the receipt.

The offer

So here is the offer, and it's the same one the page makes. Open the configurator, switch on the engines your buyers actually ask, pick the model each one answers on, choose 20, 50 or 100 questions, and read the price before you decide anything. Add a competitor's report for $50 or the 90-day action plan for $150 if you want them, or take the whole thing monthly at the same price with the plan included, and the rail shows every line either way. Pay, approve the questions we suggest, and the engines start within a minute; the report opens on your dashboard and is emailed the moment it lands, with every number on it traced to a stored answer you can read yourself.

If you'd rather we made the changes the plan calls for, we implement it at our agency rates, quoted by conversation rather than as a line on the report, and the proof that it moved arrives in the next report, on the same questions, with the model named on every answer. We look forward to growing with you, and our agents are at your beck and call.