The Journal
What we learn running grounded AI audits: the answers, the costs, the mistakes, and the work that comes next.
We pay retail for every grounded answer, across five AI engines, and read the bill line by line. The lesson: a model's per-token price tells you almost nothing, because the model decides how much of your money to spend on each question.
Read The Story ↗We audited 160 of our own failed engine calls. 64% were us: credit and quota, not the engines. Here is what we built so a vendor's outage can never be sold to you as your invisibility.
Read The Story ↗ChatGPT is not one thing. It is a family of models with measurably different search habits, and a visibility score that will not name the model behind it is not a measurement. It is a vibe with a decimal point.
Read The Story ↗Search gave buyers ten links and wished them luck. Answer engines give them two or three names and a verdict. You are in that group, or you do not exist for the buyers who are ready to buy today.
Read The Story ↗Ask a model with web search off and you learn what it memorized in training, years ago. Ask with search on and you learn what a buyer sees today. Confusing the two is the original sin of AEO measurement, and most of the category is still committing it.
Read The Story ↗When ChatGPT names three brands and stops, that verdict has an anatomy. Learn it and “why not us?” stops being a mystery and becomes a work item with a deadline. There are exactly five ways to lose, and all five are fixable.
Read The Story ↗About half of U.S. adults use an AI chatbot, 52% of U.S. workers use AI on the job, and four in ten adults use a chatbot to search for information. The audience moved. The question is whether the machine mentions you when it answers them.
Read The Story ↗Younger, better educated, higher earning, and clustered in a handful of places. The demographic data on AI users reads like a description of the customer every sales team wants, and that customer now takes advice from an answer engine.
Read The Story ↗Half of Americans are more concerned than excited about AI, and half of them use it anyway. That contradiction is the most important fact in answer-engine marketing, because a skeptical user reads the citations, and the citations are where you win or lose.
Read The Story ↗We put a per-engine, per-model price on the page and made the server re-derive it from the same catalogue at checkout, so the number under the button is the number on the receipt. Here is why we chose to sell that way, and what it cost us.
Read The Story ↗Five engines, the model version each answers on, 20, 50 or 100 buyer questions, any competitor's report and a 90-day plan: you pick the lines, the page prices them, and nothing runs that you didn't ask for.
Read The Story ↗The white-label custom report: your agency's name and colours on the cover, our five engines and our methodology underneath, ordered per client in exactly the shape that client needs, and re-run monthly with the plan included if you want it.
Read The Story ↗A custom report taken monthly costs exactly what it costs once, and the 90-day action plan that is a $150 add-on on a single report comes included and refreshed every month. No discount. Here is why we priced it that way.
Read The Story ↗The failed call you're refunded for, the $2 line that says what it is, the model named on the invoice, and rounding you can read in advance. The bill is part of the measurement, so we built it to the same standard.
Read The Story ↗Every line on the configurator is priced by one rule: what a grounded answer costs us on that model, divided by one fixed share, the same share for GPT-6 Astra as for Perplexity's Sonar. So the model we'd recommend is the one that fits your buyers, never the one that pays us more.
Read The Story ↗"ChatGPT" is seven models on our page, and they don't answer alike: one search an answer or five, ten thousand tokens of context or a hundred thousand. You pick the version your buyers are served, we name it on every stored answer, and the price follows the choice.
Read The Story ↗Our data methods, in the order the report is made: questions you approve, each asked on the engines you chose with live search on, stored word for word with the model named, read once by a machine and counted the same way every time. Each step, what it costs, and what it can't show.
Read The Story ↗Name a competitor on the form and, for what your report costs, with a maximum of $50, you get their report: the same stored answers, read with their name as the subject. Nobody is asked again, nobody is contacted, and the number that tells you whether your category is still winnable is on it.
Read The Story ↗Since the 9th of September the 90-day action plan is a $150 add-on, not a line folded into every report. Buy it at checkout, buy it later from the report, take it included on any monthly, or leave it out and pay for the measurement alone. Here is why we unbundled it.
Read The Story ↗Your Next Step
Your buyers are asking. Let's see what they're hearing.
Start Free Let's Talk ↗