Our Lead Magnet Is a Diagnosis. It's Not Allowed to Lie.

You've been on the receiving end of this funnel. Enter your URL, get a free report: a website grade, an SEO audit, a security scan. The score is alarming, the fix is a demo call, and somewhere in the back of your mind a fair question forms: did they make the number scary on purpose?

Usually, in some measure, yes. Give the diagnosis away, sell the treatment: it's the oldest funnel in software, and it persists because it's genuinely good. The prospect gets real information for free, and the vendor earns the conversation by being useful before being paid. But the pattern carries a structural temptation. The scarier the grade, the better the conversion. Nobody audits a free report — no procurement, no review cycle, just a number on a screen engineered by the party selling the fix. Every incentive points one direction: shade the diagnosis toward fear.

We run this funnel. Fama, our AI-visibility product, gives away a free grade: hand it a brand URL and an email, and it puts your brand through the AI engines your buyers actually ask (ChatGPT, Perplexity, Claude, Gemini, and the rest), then sends back a score out of 100, a row of checks and crosses per engine, and one verbatim AI answer, quoted on the page, from a real buying question your brand should have appeared in.

That featured answer is chosen by a rule whose written description doesn't pretend to be neutral. It says to pick "the single most striking result to feature on the public hero" and to "prefer a query the brand is ABSENT from (the gut-punch)."

Gut-punch. Our word, in our own rulebook. This is a funnel and we wrote it like one. Then the same rule finishes the sentence: if the brand is mentioned everywhere, "fall back to its weakest mention (lowest prominence) framed honestly — never fabricate an 'absent' hook."

That one rule is the whole essay. A free grader is a sales instrument, and pretending otherwise is silly. And the grade must be real anyway, enforced by the product itself, because a lead magnet that lies about the diagnosis poisons the treatment relationship it exists to start. Whether you're evaluating graders or running one, the machinery for holding both is the useful part.

Why we couldn't lie even if we wanted to

Two things keep our version honest, and the first is a category accident you should notice as a buyer: our grader's subject is checkable. We measure what AI engines say about your brand, and you can open ChatGPT yourself, tonight, for free, and ask. A fear-inflated grade in this category doesn't just risk being wrong; it risks being caught by the exact buyer curious enough to be worth having.

The second is structural, and it applies to every diagnosis-and-treatment business: our diagnosis and our treatment are the same instrument. When a free-report brand converts, the number that scared them doesn't retire. The paid product tracks the same score, weekly, using the same scoring logic. The free headline grade and the paid drill-down are drawn from one shared calculation, so there is no "marketing score" and "product score" that could quietly diverge. If the free grade were theater, week one of the paid relationship would be the reveal.

The same-instrument rule goes further than sharing a formula. The questions that feed the free score are the identical set that feeds the paid score, on every tier. Paying adds separate coverage questions that are never scored, so an upgrade can't move the line you were shown when you were a stranger. The free number is the real number, and it stays the real number after you convert.

The number itself is built to be argued with. No AI grades another AI. A rules-based reader looks at where your name lands in each engine's answer and gives it a score from 0 to 5: not mentioned at all; only in a citation list at the bottom; named in passing, mid-answer; named as a real option in the body; named in the engine's first sentence; named first and called a top pick or best. Those per-engine scores blend into the 0 to 100 grade. So every point on your report has a plain reason behind it, "a 4 because you were named in the first sentence," and if you disagree with the reason you can open the answer and read where your name sits. A grade that can explain itself is much harder to inflate than one that arrives as a black box, which is why most free graders arrive as black boxes.

That's a test you can apply to any grader that grades you: ask whether the free number and the paid number come from the same measurement. If the vendor can't answer cleanly, you've learned something about the free number.

The scare is selected, never manufactured

The fallback ladder in that rule is where funnel logic and honesty walls actually meet, so walk it.

The competitor section runs the same discipline at section level. When AI answers skipped you and named rivals, the report says so, with counts. When there's no competitor data, or you're ahead, the section doesn't soften its wording. It removes itself, rather than show an awkward "0 competitors beating you." A scare section that deletes itself when there's nothing to be scared of is rarer than it should be.

The same care runs one layer down. When you're absent from a question, the report looks at who the engines cited instead of you, and it keeps two findings apart: someone else was cited, versus nobody was cited at all. Those are different problems with different fixes, and a scare that blurred them would send you chasing a rival who isn't there.

Strictest of all: the section comparing what AI says about you against your own brand record refuses to issue verdicts to strangers entirely. The rule draws the line: Fama "never asserts falsity to a stranger (that would be the max-damage failure for a precision product)." To a stranger it only shows the discrepancy and asks: here's what AI says, here's what your record says, you tell us which is wrong.

There's a quieter honesty problem on the other side, and it's the one most graders never think about: accidental deflation. A grader that reads its own technical hiccups as your invisibility is lying just as hard as one that fabricates a scare; it just gets to feel innocent about it. So the scoring distinguishes an engine we never got answers from this run (marked not measured) from an engine that answered without you, because "'we never asked it' differs from 'we asked and the brand wasn't there.'" The overall score is recalculated over only the engines that actually answered, and an incomplete run is labeled incomplete instead of getting a silently worse number. Our flaky day is our problem, not your score. You can see the distinction on the report itself. The row of engines under the grade is a row of checks and crosses, and hovering any engine tells you what the mark means: mentioned in three of ten answers, say, or simply not measured. A cross that means "we couldn't ask" is never allowed to look like a cross that means "we asked and you weren't there."

The report also admits its own age. A grade computed today carries no label; one computed yesterday says "audited yesterday," and older ones count the days. That is a small thing, and it costs conversions, because an old scary number is still a scary number. But a diagnosis that hides when it was taken is a diagnosis you can't check.

The treatment is not allowed to touch the diagnosis

This is the mechanism I'd defend in a fight.

The treatment Fama sells, at its core, is truth maintenance: confirm the facts about your brand, correct what AI gets wrong, feed grounded material to the sources engines read. The obvious cheat sits right there: the moment a customer starts paying and confirming facts, nudge the score up. Reward the behavior. Every gamified product does it. The customer feels progress, renews, tells a friend.

The scoring logic carries a written rule that exists to kill that idea permanently: "The Brand IQ integer is PROVABLY immune to the verified-facts ledger." The score reads what the engines actually said, and nothing else. "Confirming/correcting a verified fact CANNOT move Brand IQ," the rule continues, and then it addresses future engineers directly: "DO NOT add a verified-facts / fidelity parameter to this function." An automated check enforces it: it computes the score with and without confirmed facts present and refuses to pass unless the number is identical.

One thing moves the score: the engines changing their answers, captured by the next measurement. If the treatment works, the instrument sees it, out in the world, not in records we control. If it doesn't work, the instrument says so, to a paying customer, on a chart we built. That is a terrifying thing to sell. It's also the only version worth selling, because it's the only version where the number going up means anything. And the instrument only looks when you tell it to. Monitoring is off by default; you choose weekly or daily re-audits, and once the line is on, a score shift or a newly wrong claim arrives as an alert rather than a dashboard you have to remember to open.

What the honest funnel costs, and what it buys

The costs land exactly where a growth team would feel them. A brand that grades well doesn't convert on fear: their report features an honestly-framed weak mention, or nothing, and the competitor section has removed itself. Our funnel is structurally worse at scaring the healthy than a dishonest one would be, and a competitor's grader can always out-scare us, because fabricated urgency has no supply constraint.

If you run lead magnets for your own brand, the same four walls translate directly.

And this is what the honest version buys. The free grade is the first claim your company ever makes to that buyer — before the demo, before the sales call, before any human speaks. If that first claim is theater, you've taught them your numbers are negotiable, and you now get to sell them a measurement product. When the diagnosis is real, the treatment inherits the diagnosis's credibility: the scary number that brought them in and the weekly number that keeps them are the same number, from the same calculation, that refused to lie when lying would have converted better.

The gut-punch is real. That's why it lands.