The playbook behind The Citation Lab.

A document written for someone else

The Citation Lab runs on AuditSpark's audit engine — the same tooling that scores the free check at /check/ and the hero panel on our home page. Before we were a customer of that tooling, we read the material AuditSpark had written to help other agencies decide whether to build a business around it at all. One of those documents, the sixth in a set of six, was called "The Agency Playbook for GEO Services." It wasn't written for us. It was written for any marketing shop wondering whether AI-visibility auditing was a real service line or a fad, and if real, how to structure and price it.

We read it the way you read a competitor's homework: skeptically, looking for the seams. There weren't many. The paper's central observation was simple and, by the time we read it, already visibly true — the web had spent two decades competing to rank, and was now starting to compete to be cited, inside an answer an AI engine writes and hands to someone who never has to click through. Most businesses were failing at that for boring, checkable reasons: crawlers quietly blocked, content that reads fine to a person but can't be lifted into a quoted answer, no independent evidence anywhere online that the business is who it says it is, and — more often than any of the above — nobody had ever actually checked. The old "we'll improve your rankings" pitch was getting harder to differentiate on, at the exact moment a new, harder-to-answer question was showing up on prospects' minds. That's a timing observation, not a sales trick, and it's the one that convinced us there was a real business here, not just a plausible-sounding one.

What struck us on a second read wasn't the market-timing argument — that part was easy to believe, because we were living it ourselves as a small studio wondering what to build next. It was how little of the paper was about marketing at all. Most of it was structural: how to price a fixed-fee report without either underselling the work or padding it, how to keep fix-work from turning into an open-ended engagement that quietly resents its own client, when a recurring retainer is honest and when it's just a name for an invoice nobody asked for. Those are the questions that make or break a services business long after the pitch has landed, and they're the ones most "should we get into AI SEO" conversations skip past entirely.

The ladder we didn't design

The paper's practical core was a four-step service ladder, argued from first principles rather than handed down as a template. Start with a readiness check that costs the prospect nothing — not as a loss leader in the usual sense, but because a specific, checkable finding sells itself in a way a vague pitch never does. Follow it with a paid diagnostic report: a fixed fee, priced against the value of the decision it informs rather than the hours it took to produce, delivering a prioritized, structured list of what's actually wrong. Then scoped fix-work — one discipline at a time, fixed-fee, a defined deliverable, never an open hourly meter. And, only for clients who want it, a monitoring retainer, justified by a genuine technical fact rather than invented to pad recurring revenue: engine behavior actually drifts over time, so a snapshot from six months ago can quietly mislead someone who's still relying on it.

We didn't set out to copy that ladder. We set out to build a business, tried a few shapes, and kept ending up back at something close to it — because the reasoning underneath each rung held up regardless of who was proposing it. What's on this site today is, tier for tier, close to verbatim: the free readiness check lives at /check/, no email gate required to see a result. The Diagnostic report, Fix work, and Monitoring tiers on /pricing/ are the paid three rungs, with the same shape the paper argued for — fixed fee for the diagnostic, scoped and fixed fee for fix-work, monthly for monitoring. That's not a coincidence we're glossing over. It's the whole reason this page exists: you can click through to both pages right now and check the claim yourself, which is very much in the spirit of a site whose entire argument is that evidence beats assertion.

Sell the finding, not the fear

The paper had a phrase for the first rung that stuck with us: sell the finding, not the fear. It's easy to build an AI-anxiety pitch — "AI search is changing everything, you need to act now" — and plenty of agencies do. The paper argued that approach ages badly and converts worse than just handing someone a real, specific result: here is what's blocking your site, here is who's showing up instead of you. A finding is checkable. Fear is just a mood, and moods fade by the second follow-up email.

That's the entire reasoning behind why the check at /check/ costs nothing and comes with no catch. It's also, not coincidentally, About principle 02 on this site: "Free diagnosis." "We give the diagnosis away and charge for the cure. You should never pay to find out whether something is broken" isn't a marketing decision we made independently of the paper — it's the paper's own argument, restated as a promise to the people using it.

The line we took most seriously

The single hardest sentence in the paper wasn't about pricing at all. It was a warning, and it's worth quoting the spirit of it directly: nobody can honestly guarantee that any AI engine will cite or recommend a given business. Engine behavior is proprietary, it isn't published, and it changes without notice. What an agency can honestly offer is narrower and less exciting to say out loud — remove the technical barriers, improve the structural and evidential quality of the site, and measure position over time. The paper listed "guaranteeing citations or recommendations" as the single most damaging mistake a business in this category can make, ahead of seven others, because the audience for this work is already primed to be skeptical of AI hype, and that trust, once spent on an overclaim, doesn't come back.

That line is the direct ancestor of this site's own editorial position. About principle 01 — "No guarantees" — and the Home page's "Our position: No promises. Evidence." band aren't house style we invented independently and later noticed lined up with the source material. They're inherited from it, almost word for word in spirit if not in phrasing. We didn't discover on our own that overpromising was fatal in this category. We read someone else's hard-won warning about it, believed it, and built the rule into the site before we'd sold a single engagement.

Why this is worth saying out loud

None of this makes The Citation Lab an AuditSpark product with our name on the door — the site, the copy, the pricing, the judgment calls about what to publish and what to hold back are ours, and so is every mistake in them. But it would be a small dishonesty to tell this story as if the shape of the business arrived to us by original insight, when a good part of it arrived by reading someone else's document closely and taking its argument seriously — including, especially, the part of that argument that told us to promise less than we might have wanted to. A site whose whole premise is that evidence beats assertion should be able to say plainly where its own structure came from. This is where it came from.

It's also why the rest of this Resources section exists in the shape it does. The five technical guides sitting alongside this page walk through the disciplines the paper only gestured at — crawler access, answer-shaped content, how a citation actually gets measured — because the paper argued a business like this earns trust by teaching, not by withholding what it knows until a contract is signed. This page is the one exception in that section: not a guide to a discipline, but the account of where the whole approach came from, told once, so it doesn't have to be repeated on every other page that quietly assumes it.