The project write-up is the highest-return page you own

The single highest-return page a medium-sized business can publish is a properly detailed write-up of a job it actually did, because it is the only content on the site that a language model could not have produced without you.

That sentence used to be a nice-to-have argument about trust. It is now a distribution argument. Retrieval systems select for pages that contain something the model does not already know, and a page describing a real project on a real site with real constraints is exactly that. Everything else on your site is, in information terms, a restatement.

The PR side has already moved. A 2026 survey of digital PR practitioners put 95.9% of campaigns on data-led or original content. Journalists stopped taking anything else years ago. What changed is that search now behaves the same way.

Why the generic version fails

Google's own generative AI guidance names the content type it does not want, and the example it gives is something like 7 Tips for First-Time Homebuyers. That is not a swipe at listicles specifically. It is a swipe at pages assembled from what is already public.

A case study written from the brochure is the same failure in a different costume. A sentence saying we delivered a high-quality fit-out on time and on budget for a leading client in the healthcare sector contains no information. It cannot be quoted, cannot be checked, and cannot be distinguished from a hundred identical paragraphs.

The version that works contains the things you would only know by being there.

What goes in a write-up that earns its place

Six things, and most companies have all six sitting in a project manager's inbox.

The constraint. Not the brief, the constraint. A live tenant, a heritage overlay, a 4am to 6am work window, a slab that turned out to be 40mm out of level, a lead time that blew out and forced a substitution.

The decision and the alternative you rejected. This is the most valuable paragraph on the page and the one companies skip. We specified X, we had considered Y which was cheaper up front, and we ruled it out because of Z. That paragraph gets quoted in sales conversations for years.

The numbers you own. Square metres, storeys, duration in weeks, number of trades on site, tonnes, litres, units, defects at handover. Not results you cannot verify. Facts about the work.

The thing that went wrong. One paragraph. It is the strongest credibility signal available to you and almost nobody uses it, because sign-off culture strips it out. A write-up that admits a two-week delay caused by a supplier and explains how it was absorbed reads as true. The flawless version reads as marketing.

Names, where you have permission. The suburb, the standard, the product code, the certification, the software. Specificity is what makes a passage retrievable.

Photography of the actual job. Phone photos from the site, dated, are worth more than a rendered visual. They also cannot be generated.

The permission problem, solved properly

The usual blocker is that the client will not let you name them. Fine. Almost none of the value is in the name.

Write it as a 14-storey commercial tower in North Sydney and keep every other detail. The specificity that matters is technical, not nominal. Get sign-off on the technical content, which clients approve far more readily than a logo on your site, and publish.

Build the permission into the contract instead of asking after handover. One clause, granting the right to publish a non-identifying technical write-up and site photography, agreed when everyone is happy at the start rather than negotiated when they have moved on.

The production model that survives contact with a busy business

Case studies die because they are structured as a writing task assigned to someone with no first-hand knowledge. Invert it.

Book 30 minutes with the project lead. Record it. Ask six questions: what was the constraint, what did you decide and what did you reject, what are the numbers, what went wrong, what would you do differently, what did the client keep asking about. Do not ask them to write anything, ever. Writing is the bottleneck, not knowledge.

Transcribe. Draft from the transcript, keeping their actual phrasing wherever it is clearer than yours, which is more often than you expect. Send the draft back for a technical accuracy check only, with a clear instruction that you are not looking for polish. Publish.

That is a two-hour cycle per case study, of which the expert's time is 30 minutes. A firm doing 40 projects a year can publish 20 write-ups. Most publish two.

What they actually pay for

Not traffic, mostly. Case study pages rarely pull big search volume, and if you judge them on sessions you will kill the programme in a quarter.

They pay in three other places. They get cited, because they contain retrievable facts nothing else on the web has. They close deals, because sales sends them mid-cycle to answer the have you done this before question with evidence instead of assertion. And they feed everything else: the LinkedIn post, the capability statement, the tender response, the PR pitch that actually has a story attached.

Judge them on assisted revenue and on citations, not on organic sessions.

The honest cost

Twenty write-ups a year is roughly 40 hours of expert time and 60 hours of production. That is real money and it competes with billable work. The programme fails at the sign-off stage more often than the writing stage, so if legal or the client-facing lead will not release technical detail, do not start it. You will produce twenty pages of nothing and conclude that content does not work.

It is also slow. Citations accumulate over quarters. The sales-enablement value shows up almost immediately, which is usually what keeps the programme funded long enough for the search value to arrive.

Everything else on your site, someone else could write. Only you were on that site.