Claims you can
actually defend.

They want to not fail the next audit. Jobs-to-be-done is the habit of asking what a buyer is really trying to fix, rather than what they are shopping for — and the answer usually changes who you are competing against.

• Verity thought it had six competitors, all training platforms. On the job the buyer actually has, its real rival is the consultant a company calls after a failed audit, at $180K, and the spreadsheet they are already using, at nothing.
• The language bank is evidence, not invention. Customers say “we got dinged on our training records”. Verity’s homepage says “comprehensive compliance enablement”. Nobody has ever typed the second phrase into a search box.
• The trigger is a moment, not a condition. Not “companies need compliance” but “an auditor found a gap eleven days ago”. That is when budget appears.

Every finding is tagged Evidence-Based, Inference or Assumption. The prompt refuses to invent a single customer phrase.

Crowded language is the term for a claim so widely used it no longer carries information. A buyer reading six sites in one afternoon cannot separate them — so they separate on price, which is the conversation Verity keeps losing.

• The competitive frame is the set of alternatives a buyer holds in their head when comparing. It is often not the set you think, and it decides what you have to argue against.
• White space has to clear three tests: true of us, unclaimed by anyone else, and something a buyer actually cares about. Two out of three is not an opening. Only one Verity claim survived all three.
• The complaint pattern is the tell. Across every rival’s one-star reviews, the same failure repeats — reporting breaks when an auditor asks for it. A category-wide complaint is an invitation.

The prompt also names the ground Verity cannot take. Someone already owns cheapest, and attacking it would just spend the budget.

Segmentation, targeting and positioning is one decision in three steps: cut the market into groups, choose which to serve, then say what you are to them. Most companies do the first two and skip the third, or do all three without giving anything up — which is not positioning, it is a description.

• The part that gets skipped: naming who you are not for. Verity gives up enterprise, who build their own, and companies under fifty, who cannot clear the price floor. Together that is 38% of current pipeline.
• Segments defined by situation, not size. “Mid-market” is a demographic. “Mid-market with an audit finding in the last twelve months” is a segment with a deadline and a budget that already exists.

The first two positioning statements failed the competitor test — a rival could have said them without lying. The prompt shows both rejected versions and why, rather than only the one that worked.

Points of parity are the things a buyer expects. Having them wins nothing; missing one disqualifies you. Security certification lives here — being certified never won a deal, and not being certified has lost plenty.

Points of difference are the reason someone picks you over an adequate alternative. Most marketing confuses the two and spends the whole budget shouting about parity.

• The test that separates them: could a named competitor put this exact sentence on their own site without lying? Verity’s “engaging content employees actually finish” survives about four seconds of that.
• The proof audit grades each survivor PROVEN, CLAIMED or [NEEDS PROOF] — and the prompt will not invent a statistic or a customer name to close a gap. It says what evidence is missing and the cheapest way to get it.

Seven of nine claims moved to the stop-saying list. That list being longer than the keep list is the normal result, and the prompt says so rather than softening it.

Without it, every writer invents their own value proposition and the market never builds a single clear idea of what you do. A hierarchy is the map everyone writes from: one master message, three supporting pillars, and evidence beneath each.

• The pillars have to be MECE — mutually exclusive, collectively exhaustive. No overlap, no gaps. Overlapping pillars are the same argument twice, which reads as padding.
• Proof is ranked, not assumed: measured results beat named customers, which beat third-party validation, which beats a demo, which beats a logo wall.
• The gap is named, not written around. Verity’s third pillar has no evidence yet, so it is marked and does not ship. Writing persuasively around a hole is how unprovable claims reach a customer.

The whole hierarchy then compresses into one sentence a customer could repeat accurately to a colleague. If it cannot, it is not finished.

“Professional yet approachable” constrains nothing and describes every company on earth. A usable voice is stated as trade-offs — plain over clever, specific over comprehensive — because a trade-off can be applied to a sentence and an adjective cannot.

• The de-slop pass is the part nobody else has. Machine-written copy has tells: opening on a grand abstraction, empty verbs like unlock and elevate, the rule of three used as a tic, the “not just X, but Y” construction, and closing paragraphs that summarise what was just said.
• Verity’s headline carried four of them in eleven words and said nothing a competitor could not also say.
• The read-aloud test catches the rest. Any sentence you cannot speak in one breath without sounding like a press release gets rewritten.

There is something useful about an AI prompt whose job includes telling you how not to sound like an AI. The tells are predictable, which is exactly why they can be listed and removed.

A campaign is not a set of assets. It is an attempt to move a specific group of people from one belief to another. Briefs that skip that produce activity and no result, which is most briefs.

• The honest version of “what they believe now” is the hardest line to write, because it is usually unflattering. Verity’s buyers think every vendor in the category is interchangeable. They are largely right, which is the problem.
• The communication objective has to be one thing. Awareness, comprehension, preference, trial or action. Most briefs claim all five, which means the campaign is optimised for none. Verity does not have an awareness problem; it has a comprehension problem.
• The insight is an observation, not a statistic. Nobody is proud of buying compliance training. They are frightened of the audit. Everything in the campaign follows from that sentence.

The brief also states what result would mean it failed — before it runs, while that is still an honest number rather than a negotiation.

Every ad variant changes exactly one element from the control — the hook, the proof, or the call to action — and says which. Change three and a winning variant tells you nothing about why it won, which is how teams run tests for a year and learn nothing.

• The traceability table is the control that matters. Every headline and claim is mapped to its pillar and its evidence. Any row without proof is marked and does not ship. Two of Verity’s did not ship.
• Findable by AI assistants is a real requirement now. Buyers increasingly ask Claude or ChatGPT rather than a search box, and an assistant quotes a sentence out of context. So the claim goes in one complete sentence, the evidence sits in the sentence beside it, and the company is named rather than called “we”.

The prompt finishes by running the de-slop list over its own output and reporting what it caught. Marking your own homework is imperfect, but it catches the obvious cases.

The prompt works through six layers in order and commits to one answer. Reach, relevance, comprehension, credibility, friction, fit. Listing all six as problems is what a report does; naming the binding one is what an analyst does.

• Verity’s buyers understand the claim and do not believe it, because nothing on the site proves it. That is credibility, and it is fixed with evidence rather than more traffic — which is where the budget was going.
• The cost is computed, not asserted: $2.1M of pipeline a quarter, with the arithmetic shown, so the fix can be prioritised against everything else competing for the same week.
• A hypothesis has to be falsifiable. One variable, a minimum lift worth acting on, the traffic needed before the number means anything, and a kill criterion agreed in advance. Without the kill criterion a failing test just runs longer.

It also names what to stop doing. An analysis that only adds work has not finished.

Most positioning work ends as a sixty-slide deck that three people have read. A three-page brief a freelancer can act on is worth more than all of it, because the test of brand work is whether someone else can produce on-brand work from it without asking you a question.

• Guardrails are the section nobody writes. Not what we say — what we will never say, and why. No “engaging”. No implication that a regulator has approved the product. No customer named without written permission.
• The review trigger stops it going quietly stale. A competitor move, a new segment, a product change — named in advance, so the brief gets revisited deliberately rather than being quietly ignored for three years.

If a section cannot be completed because the upstream work is missing, the prompt says which one to run rather than filling the hole with something generic. A brief with an honest gap is more useful than a complete one that is partly invented.