We audited our own lead magnet.
Here's what we cut.

The 9-Step AI Funnel Audit is the free PDF we hand people when they ask how to grade their own funnel. Nine checks, an A–F rubric for each, before-and-after fixes, and an ICE score at the end so you know which fix to do first. It's the most useful thing we give away.

Last week we ran it on itself. The lead magnet, the /playbook page that delivers it, and the follow-up that's supposed to come after. Same rubric we'd use on a client.

It failed check 03 before we got past the cover.

Check 03 failed first: proof we couldn't point to

Check 03 in the audit is proof density. The PDF says it plainly: if a client won't be named, don't use the quote, because "anonymized proof reads as fabricated." Then you turn back to the cover, and the cover says "12,400+ readers" and "★ 4.9 · 1,800 reviews."

That's a free PDF. There's no review widget on it, no store listing, no place where 1,800 people could have left a star rating. When we went looking for a source for either number, there wasn't one. Same for a "Forbes Contributor" credit pill on the landing page.

Then the reviews section. The /playbook page had three reader quotes praising the PDF, one of them with a tidy "18% lift in two weeks" result attached. The page history shows those quotes were added on May 14 by an automated site rebuild. The PDF file itself was created on May 17. Nobody reviews a document three days before it exists.

None of this was written to deceive anyone. It's what happens when templates, page builders and "make this section stronger" passes stack up over time with nobody checking the receipts. But a buyer doesn't see the process. They see a free PDF with 1,800 reviews and do the math.

What we did: one test for every claim. Can we point to where this number came from? If yes, it stays. If no, it goes. The cover now says how many pages it has and nothing else. The reviews section is gone.

The stats that sounded like research

The second pass was inside the PDF. Three numbers had the shape of research and none of the citations:

• Check 05 said operators who miss the second and third follow-up touches "lose 53% of qualified intent" in those gaps.
• Check 02 said a benefit-first CTA beats "Book a call" by "1.4×–2.1×" across a stated number of our deployments.
• The retention fix promised to recover "20–30% of would-be churn", "typically a 6-figure annual win."

Here's the thing: the advice around those numbers was right. Follow-up windows do leak qualified leads. We wrote a whole issue on it, Speed to Lead (#157). The number wasn't doing any work the mechanism wasn't already doing. It was just making the sentence a liability.

So check 05 now reads: "Most agencies miss touches 2 and 3 — and that gap is where most qualified intent leaks away." Same instruction. Nothing to defend.

If a stat in your copy can't survive someone replying "source?", it's costing you more trust than it's earning.

The math that was loose

Check 08 is retention. The old version said a one-point churn reduction "compounds to 12 points of LTV." That's the kind of line that sounds right until someone runs it. So we ran it:

// average customer lifetime ≈ 1 / monthly churn
churn 5%  →  1 / 0.05  =  20 months
churn 4%  →  1 / 0.04  =  25 months

// one point of churn = 25% longer lifetime = 25% more LTV

The real number is better than the vague one, and anyone can check it with a calculator. That's the standard: if the claim is arithmetic, show the arithmetic.

The small breaks that say "nobody checked"

None of these would cost you a lead on their own. Together they tell a careful reader the thing hasn't been opened in a while:

• The page count was wrong. The landing page said 22 pages. The file had 27. The rebuilt version is 25, and the printed page numbers now match the physical pages.
• The cover title overlapped its subtitle, so the headline rendered as something like "AIweFunnel Audit."
• Three pages overflowed, with text running off the bottom.
• The CTA said "Apply for Q3 2026." It's Q4. It now just says "Apply."

Dates and counts are the cheapest thing to get right and the first thing to rot. If your lead magnet has a quarter, a year or a page count in it, put a reminder on your calendar to check it.

The promise vs. the plumbing

The PDF makes a promise: no sales sequence. You get the PDF, one follow-up, then only The Scale Brief. That's a good promise. It's also easy to break the first time someone decides playbook leads need "nurturing."

So we built the follow-up to match the promise instead of the other way round. One short note a couple of days after delivery asking how your scorecard came out, with an offer to look at it if you reply. Then the newsletter. The apply link only shows up in a P.S.

If your lead magnet promises one thing and your automation does another, the automation wins and the promise becomes the lie. Check what your workflow actually sends, not what the page says it sends. (Related: Your Form Looks Like It Works (#145) on making sure the submission reaches the CRM at all.)

What we added

Cutting was most of the work, but the audit flagged two gaps worth filling:

1. A scorecard you can fill in. The old PDF had nothing to write on. Now there's a one-page sheet: nine checks, A–F for each, plus an ICE calculator to rank the fixes. It also gives that one follow-up email a real reason to exist. "How did your scorecard come out?" is a question worth answering.
2. A page for local service businesses. The retention check was all churn and MRR, which is no help to a roofer or a plumber. There's now a page that runs the same nine checks for a trades or local service business.

Run the same test on yours

You don't need our PDF for this part. Open your own lead magnet and its landing page and answer five questions:

1. For every number: can you point to where it came from? If not, cut it or replace it with the mechanism.
2. For every testimonial: can you name the person and the date? Was it given after the thing existed?
3. Do the counts match? Pages, steps, templates, minutes. Open the file and count.
4. Is anything dated? Quarters, years, "new for 2025."
5. Does the follow-up match the promise? Read the workflow, not the copy.

Most lead magnets fail at least one. Ours failed all five.

The one-line summary

Run your own audit on your own lead magnet before a prospect does. If you can't point to the source, cut it. Pairs with #157 on follow-up windows and #151 on keeping the source attached to the lead.

P.S. If you'd rather have us run the nine checks on your funnel with you, apply here. We'll tell you what we'd cut first.