Every step below says what to do, what you should see, and why it matters. If what you see doesn't match what's written, that's a bug — write down the step number and move on.
The one thing we need judged: step 5. Everything else is plumbing. Step 5 is whether the writing finally knows what a coded magnet is.
Before you start
Two warnings. The three magnet stores take real payments — browse them, add to cart, but do not complete a checkout unless you actually want a magnet kit shipped. And in step 6 you will change a live web page on purpose; there's a one-click undo, and you'll use it.
Prefer to watch first? The walkthrough below runs 2 minutes 52 — same running order as the nine steps. It's a public product video, so it's written for anyone, not just you.
You should seeThe app itself — a dashboard. Not an error, not a sign-up page, not a price list.
WhyLast week this sent you to a page that refused you. If you get anything that looks like an error, stop and tell us — nothing below will work.
2
Your three stores
Do: look at the list. The My sites filter is already switched on.
You should seeExactly three: ferrospring.com, codedspring.com, polymagnetics.com. Click All if you ever want the whole network.
WhyYou were previously dropped into 249 sites, almost none of them yours. Now it opens on your work.
3
Open ferrospring and read the score
Do: click the ferrospring.com row.
You should seeA panel: 60 / 100, five discipline bars, and “what the engine read” — title length, word count, headings, sitemap URLs, schema blocks, links. Its top three problems are Breadcrumb schema, Homepage depth, and AI answer identity (sameAs).
WhyEvery number traces to something actually on the page. 51 checks ran; 27 passed, 13 warned, 11 failed. If a number appears with no evidence behind it, that's a bug worth reporting.
4
★ Industry Intel
Do: click ★ Industry Intel in the left menu.
You should seeThe competitive field — K&J Magnetics, Eclipse, Bunting, Master Magnetics, Dexter, Apex. Correlated Magnetics Research listed as the source worth citing. The behaviours buyers care about: align, latch, spring, hold, detent, twist-release. Plus current industry notes.
WhyThis is the engine's knowledge of your market, refreshed weekly. It's also what the writing in step 5 is built on. Tell us if anything here is wrong — you know this field better than we do.
5
Read one content brief, all the way through ← the real test
Do: go to Content and open “Every FerroSpring coded magnet, by category”. Read the whole thing — the outline and every direction point.
You should seeIt names your real categories (Demos & Kits, Alignment, Latch, Spring, Detent), your real FAQ questions, the actual competitors by name, and an industry angle. It should read like it has been on your website.
WhyYour verdict has been “it's generic.” This is the thing that changed. The question we need answered: could a writer take this brief and produce the article without asking you anything? If yes, we stay on the current approach. If no, we spend money on AI drafting. That's a real fork and it's your call.
6
Change a live page — then undo it
Do: go to Approvals & Setup. Find “Schema / JSON-LD to add” and click Approve. Then open ferrospring.com and view source. Then come back and click ↵ Roll back.
You should seeApproving publishes it to the live page immediately. The rollback restores the file exactly as it was — byte for byte.
WhyThis is the part no audit tool does: the engine doesn't just tell you, it changes the site, and it can always undo. Everything reversible is backed up before it's touched.
You should seeThe full 38-product Polymagnet line, real photos, working cart, a real checkout.
WhyWe checked every product's page price against the price that would actually be charged — 38 of 38 match on all three stores. Checkout is live and will charge a real card.
You should seeYour profile, a password change that asks for the current one, your role (manager), your sites, and payment methods.
WhyYou shouldn't need us to manage your own login. Manager means you can see everything and approve and ship — but you can't delete a site or change other people's accounts.
Being straight with you
What we already know is still weak — don't waste your time finding it
Four growth assets were rebuilt overnight — pricing, competitor teardown, social hooks and the lead magnet. They now use your real catalogue, your real competitors by name, your buyer pillars and your own FAQ questions. The competitor card no longer says “pick your top 3” — it names them. Worth a look, and worth telling us if any of it is wrong.
All sixteen growth cards are now grounded — the remaining eleven (A/B test, positioning, customer research, social plan, link targets, link outreach, digital PR, design brief, week plan, editorial plan, cold email) were rebuilt on Aug 3. They now read from your own audit rather than a template: the A/B test runs the CTA finding the audit actually made, the week plan is your own 30-day roadmap cut into five days, link opportunities names 36 real organisations instead of listing target “types”, and the cold email finally names the checklist it is offering. Where the engine genuinely lacks the material it now says so and tells you the one thing that fixes it, instead of padding.
Still weak, and we know it: the 90-day content calendar is itself templated — every site gets “The complete guide to {your field} for your area.” The editorial plan faithfully reproduces it, so that is the next thing to fix, not the plan.
The real polymagnet.com still can't be audited. Shopify rate-limits us. That's exactly why you're testing on the three clones.
polymagnetics.com scores lower (57 vs 60) than the other two. That's real — it is the only one of the three with measured traffic, and the traffic is thin. Not worse pages.
What we need from you
1. Is the brief good enough?
Today the engine produces a brief — the structure, the real questions, the competitor angle, the keyword. A person writes the article. The step up is having AI write the full draft, which costs money per article.
After step 5: could a writer run with that brief as-is?
2. What do we point at the real polymagnet.com?
Options: keep proving it on the clones · build a Shopify connection so approvals can touch the real store · or audit it read-only and apply fixes by hand.