The Line on Your Homepage Nobody Believes, and the AI Audit That Finds It

There is a line on your homepage nobody believes

You know which one it is.

It is the sentence you fought for in the review. The one with "AI" in it. You paid someone to write it, you approved it, and your buyer reads it, nods, and scrolls. They read the same line on four other sites before lunch.

Here is the whole article in one move: this is not a writing problem. It is a claim problem. You wrote a claim that cannot be argued with, and a claim that cannot be argued with cannot be believed.

A category is not a claim

"We are AI-forward" is not a claim. It is a category. There is no position inside it. No number, no distance, no before and after. A buyer cannot disagree with it, which means they cannot agree with it either. The only thing they can do with a category is skip it.

A claim on a scale is different. Put a stake in the ground at an actual point on an actual distance, and now a person can argue with you. Arguing is the first sign anyone believes the thing is real. Placing a claim on a scale rather than in a category is what makes it arguable, and arguable is what makes it useful.

I have watched companies call themselves AI-forward from step two of a distance that runs a great deal further than they think. Two steps in, and they say the word "arrival." That word is what stops the travel. It stops it internally, because the team believes the work is done. It stops it externally, because the buyer reads an arrival claim from someone standing exactly where everybody else is standing.

Arrival: the belief that "we use AI" is a destination. It is step two of a long distance, and calling it arrival is the thing that halts the walk.

The videos all came back identical

We were screening for a role. Every applicant got the same interview questions to answer on video before any live conversation, so the team could hear the person before meeting them. That was the design.

What came back was the same video.

Near verbatim, applicant after applicant. They had taken our questions, dropped them into AI, likely with the job description attached, and AI handed back the same answers to everyone. One written answer still had the AI's own closing line pasted on the bottom, offering to craft more follow-up answers.

That last detail is the one to hold onto.

Nobody catches it. The process keeps running. And now you are comparing one AI's answers to another AI's answers while believing you are reading people. The stage still works. It just quietly stopped doing the job it was built for.

The fix was not a better question. It was a different job for the tool. Put a timer on the answer, so the question is seen for the first time when it must be answered. Unrehearsed. That is the only version of the answer that tells you anything about the person in front of you.

Now look at your homepage. Same condition. If you asked AI to write your claim, you got the same claim as everyone who asked the same question. The tell is the pasted closing line: the phrase that could live on any competitor's page without anyone noticing. You did not write that. It arrived with the answer, and you published it.

The job you should be giving AI

AI is not a content mill. Treated as one, it hands you the most average possible version of whatever you asked for, and you put the average version on your homepage and then wonder why nobody believes it.

Treat it as a thinking partner and a diagnostic engine and the job changes completely. It does not write the line. It finds where you actually sit on the scale, and it tells you which of your claims you can prove.

That work runs through a sequence I use: the Agentic AI Diagnostic Sequence. Six moves, from constraint to build.

Run on the homepage line, it makes four decisions you cannot skip. Each one carries a cost, and the cost is the reason most people stop before the answer.

Step 1: find the constraint, then accept that it may not be your homepage

Step 1, Constraint (TOC). Identify the binding constraint. Theory of Constraints asks one question: where is the leverage? Is this work at, feeding, or near the binding constraint?

For the homepage line, that resolves to a binary decision.

Is your homepage claim at, feeding, or near the constraint? Yes, proceed. No, leave the line alone. Do not rewrite it. Do not touch it.

That second answer is the one that costs. You came here to fix the sentence, and the honest answer may be that the sentence is not the problem. If deals die in follow-up, if the pipeline never converts past the first call, if the offer itself is unclear, then your homepage is not the constraint and polishing the claim is motion that feels like progress. It is the most popular way to avoid the real work. If Step 1 comes back no, the discipline is to close the tab and go fix the constraint. Most people will not, because the sentence is the part they can see.

Triage: the valuable output is the deletion

Before you org-chart anything, run the front-end triage. This sits between Step 1 and Step 2. TOC asks where the leverage is. Eisenhower asks what the leverage of the work is: urgent and important goes first. Then you eliminate.

Write down every item on your homepage project. The new hero image. The "AI-powered" badge. The competitor teardown. The SEO pass. The copy refresh. The testimonial page. The brand color.

Now triage, then delete.

That is the cost, and it is real: this filter exists to delete most of the project, and the deletion is the output. The badge is already designed. The copywriter is already booked. The person who wanted the new hero image will notice it did not happen. You have to let those items sit undone and visible, on purpose, because a homepage project is a list of things people can do. Deciding which ones you will not do is the actual work, and it is the part nobody wants to be in the room for.

Step 2: find the owner before you touch the sentence

Step 2, Role (AI-Empowered Org Chart). Place the work in the org structure. One question: who owns the claim?

If a person owns it, proceed. If nobody owns it, stop. Fix ownership before rewriting. An unowned claim cannot be rewritten into a true one, because a rewrite with no author has no one with the standing to make a claim the company will stand behind.

Here is the cost of asking the question out loud. You will find out the sentence got onto your homepage because a committee liked it. Nobody defends it, because nobody wrote it. Naming an owner makes one person accountable for a claim that may turn out to be false, and that is not a paperwork step. That is a person putting their name on a number. Plenty of orgs would rather keep a bad claim than assign a responsible party, and that preference is exactly what you are looking for when you run Step 2.

Step 3: the claim stays Human, the audit goes to AI

Step 3, Work Type. Classify each piece as Strategic, Tactical, or Analytical, then assign each slot Human or AI.

Two slots matter here, and the assignment is the decision, not a preference.

Your company's claim is Strategic. That slot stays Human. And this is the discomfort: it means a named person has to write the line and sign it. You cannot tell the board that AI wrote the claim, because the whole point of a scale claim is that a human chose the position and a human stands there. You have to be willing to lose an argument about your own number.

The evidence audit is Analytical. That slot goes to AI. Read every page, proposal, call note, testimonial, and competitor page, then report which claims have proof behind them and which have a hole.

This is the part almost everyone gets backwards. They hand the strategic slot to AI, ask it to write the line, and do the analytical work by hand or not at all. Reverse the two and the whole thing works.

Step 4: give AI the authority to call your best sentence unearned

Step 4, AI Employee concept (Diagnostic Canvas). Define the worker, not the task.

Its job: tell me which of my claims are unearned. Its inputs: the site, the proposals, the notes, the testimonials, the competition. Its output: every claim you make, with the evidence attached, or an honest blank where the evidence should be sitting.

That output is the gate. No claim goes on the homepage until the audit has either attached evidence or returned a blank. A blank means you fix the claim or you cut it.

The cost is the authority. You are asking the tool to come back and tell you that your best sentence is unearned. It will point at the line you fought for in the review, the one you like, and hand it back with nothing behind it. Most people quit the audit at the first finding. They argue with the machine instead of reading the blank, and the blank was the answer.

Steps 5 and 6 are conditional, not optional

Step 5, PRD. Step 6, Build. Both are real, and both are conditional on what Step 4 returned.

If the audit reveals that the constraint needs a system, you write the requirements and you build the thing. That is a real outcome and a real sequence, and skipping to build without it is how companies buy software they cannot use.

For the homepage line, though, the decision is no build. The fix is a claim placed on a scale, and a scale claim is not software. Say that decision out loud, because if you do not, someone downstream is waiting on a build that is not coming. Steps 5 and 6 stay on the map. They are just not the next move for this claim.

The proof is in the starting line

An attendee at one of my recent US AI events, a complete beginner whose testimonial I now quote, asked whether a coffee-shop owner who does not know what AI is could even be a target for the event. Another attendee that year turned up with a brand-new computer and asked how to turn it on.

Judged on where he started, that man was the least likely person in the room to get anything out of a two-day event. The obvious target would have been the people already using AI.

Then he sent me the number.

Two days later the beginner put a number on it: a hundred thousand dollars this year. And the number did not come from the thing they built in the room. It came from the rewiring of how he thinks about solving problems.

His entire experience with AI, in his own words, was signing up for the course. That is the starting line, and it is the whole proof.

So the do-differently is two things. Do not screen beginners out of a teaching offer. And when you collect a testimonial, capture where the person started, not just what they got. The starts are already sitting in your inbox. Nobody wrote them down, because everyone was collecting endings.

Now read your homepage claim as a testimonial. "AI-powered results" is an ending with no before, which is why the buyer skips it. A starting line is more believable than anything you can say about your platform, because it is a measured distance. It has a before. It can be argued with. That is the same move you just watched a beginner make, and he made it in a sentence about a course.

Why you are not going to do this on Monday

Let us be honest about the cost.

If you place your real position on the scale, it will be smaller than your competitor's claim. They say AI-powered everything. You will say you automated one thing and it works. That feels like bringing a knife to a gunfight, and the fear is real: your buyer might read honest and hear behind.

That is the trade. You give up a claim nobody believes for one everybody can check, and checking cuts both ways. If you say you are two steps in, someone will ask what steps three through twelve are, and you had better know.

The rewiring is harder than the rewrite. You have to stop asking AI for an answer and start asking it for a diagnosis. That means handing a tool the authority to tell you your best sentence is unearned, and then not arguing with the blank. Most people will run the audit, hit the first finding, and quietly go back to asking for copy. If that is you, you have not changed the job. You have just bought a faster content mill.

And there is a second resistance, straight out of that event. You want to be seen as sophisticated. A beginner feels like weak proof. Screening for a buyer who already knows what AI is feels safer, and it screens out the person who put a hundred thousand dollars on it. Where the person started is the proof. It is not the disqualifier.

The same claim, placed two ways

What to do with this on Monday

Open your homepage. Read the line.

Ask the first question: is this claim at, feeding, or near my binding constraint? Answer it honestly, and then do one of two things. If the answer is no, close the tab and go fix the constraint. If the answer is yes, name the owner, hand the evidence audit to AI, and let it come back with a blank.

Then cut the claim or earn it. No nicer adjective. No new hero image. Nothing from the deleted list.

The line nobody believes is not really a line. It is a claim you never placed on a scale. Place it, and your buyer can finally argue with you. That argument is where the sale starts.

<!-- cluster-links:begin -->

Related reading

• AI Did Not Make You Faster. It Made Your Thinking Visible.
• The Decision You Think Only You Can Make Is Probably a Script
• Your Last Ten Mistakes Aren't Ten Mistakes. They're One Habit Wearing Different Clothes.

<!-- cluster-links:end -->