Signed off 4 September. Google row added. Section 7 (language changes) still needs Mo's numbers. Everything else on this page is the build brief for the second staging site.
Show a slice of what we hold. Hold the rest. Every held item already exists when the screen renders.
The real check gives us 4 things a manufacturer cares about: who ChatGPT named, why it chose each competitor in its own words, which page on each competitor's site earned the place, and what their homepage says next to the visitor's. The screen shows 1 of each and says what we are holding. The button is where the call is explained.
Current: a 12-second timer and an email ask. Proposed: no ask. A bar that always moves, a countdown that says how long is left, and a line that says what is happening now.

Keep this page open. The answer appears here on its own.
How the bar behaves. It starts counting down from 45 seconds the moment the question is sent, and the bar fills against that countdown, so it never sits still. The status line changes every 4 seconds: "Reading your homepage", "Working out what you make", "Asking ChatGPT the buyer's question", "ChatGPT is searching the web for suppliers", "Reading its shortlist", "Checking for your name", "Reading the reasons it gave", "Opening the pages it read", "Putting the competitors' pages next to yours". The 6 buyer questions run at the same time, so the wait is the slowest of the 6. When the last answer lands the bar jumps to full and the screen changes. If ChatGPT is slow, the countdown reaches "a few seconds left" and holds there with the bar at 95%, the status line still changing, until the answer arrives. It never shows 100% before the answer and it never stops moving.
Friction Technology, real answer at 06:57 UK today. ChatGPT put them first of 5. The blurred lines are the real reasons ChatGPT gave, blurred, not placeholders. The "vs yours" counts are how many things each competitor's page says that theirs does not, from the same homepage read.

This is the answer a buyer was given at 06:57 UK today. It can change with every check.
Based on current sources, here are the suppliers I would shortlist:
Below is just an initial scan of what we found. We can show you the deep dive for free on a 15-minute call, including fixes you can take away for free.
We asked ChatGPT all 6 this morning. Who it named first, and where you came.
Every relining enquiry goes to Industrial Brakes in Wolverhampton. ChatGPT names them in 4 of the 6 ways buyers ask. In the 1 you miss, they are first.
To write the answer at the top, ChatGPT went and read 1 page from each of the 5 companies. Here they are. 4 are your competitors' pages. 1 is yours.
ChatGPT said it would give a shorter list if it knew what the parts are for (aerospace, rail, wind, marine) and how many a year. A real buyer tells it both. The shorter list may not have you on it.
That is why you are first today.
A free 15-minute video call with Mo, the founder. Your answer and the competitors' pages open on screen. No deck, no pitch. You leave with the first thing to change.
Nothing to prepare. No obligation after.
The strip at the bottom of the frame is the booking button pinned to the bottom of the phone screen, 20px either side, 52px tall. It appears once the reader scrolls past the answer card and stays until they reach the big button. Average scroll depth on the old lander was 21% (Clarity, 3 September), so most readers never reached a button 4 screens down.
Tannlin, real answers at 09:05 UK today. The main card is now the question built from their own homepage line ("high-precision thin metal laser processing"), which is what the check asks once the size cap is fixed. 6 named, Tannlin absent. Across the 6 ways a buyer asks, Tannlin is named in 2 and first in both, and missing from 4.
The 6 competitor homepages were read the same way as before, so the gap table counts are against these 6, not the earlier 10.
Same screen as section 2: the names and the 12 of 40 are the same for every visitor, whatever they type.
Those are your own words, from your homepage. This is the answer a buyer was given at 09:05 UK today.
A shortlist worth considering:
Below is just an initial scan of what we found. We can show you the deep dive for free on a 15-minute call, including fixes you can take away for free.
We asked ChatGPT all 6 this morning. Who it named first, and where you came.
Your homepage says "high-precision laser processing of thin materials". A buyer who asks for exactly that is sent to Datum Alloys. Not to you.
To write the answer at the top, ChatGPT went and read 1 page from each of the 6 companies it named. Here they are. All 6 are your competitors' pages. It did not read a page of yours.
Your homepage says "Accreditations" and names none. It does not say where you are.
A free 15-minute video call with Mo, the founder. The 6 pages it read next to yours, the 3 changes to your homepage, and the first one to make. No deck, no pitch.
Nothing to prepare. No obligation after.
Same pinned button as section 2, with 20px either side and 52px tall.
The test for every row: it is real, a manufacturer gets it in 1 line, and what we hold is already in our hands. "Proof on the page" means their own page is quoted back to them. Nobody is sent anywhere.
| Fact | Shown on the screen | Held | Where it comes from | Proof on the page | Cost, time |
|---|---|---|---|---|---|
| The answer: who ChatGPT named, where they are, whether you are on it | All of it | The live ChatGPT call | The answer itself, timestamped | $0.004, 12 to 63s | |
| The 6 ways a buyer asks for what they make: who ChatGPT names first in each, and where they come | All 6 rows | The 6 full answers, and who took each enquiry | 6 live ChatGPT calls, run at once during the wait | The 6 rows themselves | $0.024, the slowest of the 6 |
| Why it chose each competitor, in its words | 1 reason | The other 4 | Same call, already extracted | ChatGPT's own sentence | Free |
| The pages it read to decide | The count, and how many were yours | The pages themselves, side by side | Same call, the citations | The count | Free |
| What it would need to shorten the list | Its own words | The re-run for the buyer's real question | Same call | ChatGPT's own sentence | Free |
| What your homepage told ChatGPT you make | Verbatim, on the wait screen | Their meta description | Their own sentence | Free, 0.4s | |
| Where ChatGPT sends the buyer instead: the job it gave each competitor | 1 job, verbatim, and the enquiry it takes | The other 3 (or 9), and which enquiries each takes | Same call, already extracted | ChatGPT's own sentence | Free |
| What the competitors' homepages say that yours does not: certifications, where they are, industries served, a who-we-are note for machines | Every gap, in plain words, with the competitor count | The wording they used, and the order | A plain read of each homepage, in parallel | Their own page quoted: "Accreditations" and nothing after it | Free, about 1s |
| Google's answer to the same question (added, Mo 4 Sep) | In or out of the top 10 | The list and the AI Overview | Google search, live | The position | $0.002, 6s, in parallel |
The Google row is in (Mo, 4 September). For Friction Technology this morning Google put them 1st and put Industrial Friction, the Cardiff competitor Jack named on the 2 June call, 4th. Two engines agreeing reads as proof.
Found in this morning's runs. All mine, none need your word.
Struck through is what the staging page says now. Below it, the proposed line. Items 1 to 3 assume the 40 goes.