20 years of documents were still answering
for the company as if each were current.

01What happens.

Nothing that was published has gone anywhere: old microsites, PDFs and subdomains stay live, mirrored or cached, and a document superseded 10 years ago carries no signal a machine reads as superseded. The estate is larger than the organisation's records show: across risk profiling of more than 119 million websites, 41% were unknown to the organisation's own digital teams.

02Why it matters.

A superseded indication or a withdrawn dosing schedule served as current reaches a clinician, and then a patient. A pharmaceutical company's medical affairs team controls what it publishes now. What it published in 2011 sits partly on sites and stores the company still controls and has forgotten, and partly in mirrored and partner-held copies it does not control, and AI reads both without knowing the difference.

Each superseded document the company still hosts costs it to keep, to secure and to maintain, paid quietly inside existing budgets. The misinformation it can cause is a separate matter, and it is paid when a clinician, a patient or a regulator receives the 2011 answer.

03The case.

This illustrative scenario draws on patterns observed across pharmaceutical and other corporate estates; no company is identified.

A pharmaceutical company has published for 20 years across a corporate site, a dozen product sites, regional affiliates, congress microsites and a document store that has never been pruned. Each publication was correct when it went out, and each was superseded in its turn by a later version, a label change, a new indication or a withdrawal.

Nobody took anything down, so a 2009 launch microsite is still live because the domain auto-renews, and a 2012 monograph is in the document store because the store keeps everything. A congress presentation from 2015 is mirrored on the congress organiser's site. A regional affiliate's page from 2017 was never migrated when the affiliate moved platform, and the old platform still serves it.

A clinician asks AI about the indications for one of the company's products. The answer is assembled from the 2012 monograph, the 2015 presentation and the current page, and it presents an indication the company withdrew in 2019 alongside the current ones, without distinguishing them.

The medical information team, asked about the answer, checks the current page and finds it correct. The 2009 microsite had no owner after the launch team moved on. The document store held the 2012 monograph among 4,000 other files with no version marking, and the congress mirror sits on a site the company does not control.

The estate had grown quietly, one correct publication at a time, and nobody had set up a review to read the wider reachable estate as a machine reads it.

04Exposure.

A company is held to what it has published, and 20 years of documents are 20 years of positions the company can be quoted on. Where the documents carry clinical content, the exposure is patient safety and licence position as well as the regulator's questions about why superseded indications were still being served. The accountability is the company's for material it published and left standing; whether any obligation is engaged on the facts is for its advisers.

The material the company still hosts costs it to keep, secure and maintain, and retiring a document the company controls ends those costs and stops that copy answering. Mirrored, cached and partner-held copies are addressed separately, through the parties that hold them. The exposure comes from unmanaged publication and missing version context rather than from keeping records. A complete archive with its dates and status readable to a machine is an asset; the same archive without them is the problem this case describes.

05What AI Readiness changes.

The estate was too large and too old for anyone inside it to hold in their head, and the review process checked the current page. Reading the wider reachable estate from outside, as AI reads it, AAAnow identifies the superseded clinical material still answering, on the company's own properties and in copies elsewhere. It shows the documents AI is drawing on for each question, with their dates.

The company then prioritises that material by the harm a wrong answer can do, retires or corrects what it controls, has partners address the copies they hold, and marks what remains with the dates and status a machine reads. The position is placed on the AI Readiness Maturity Scale™ and measured again as the material is brought under control.

Your team makes the changes; AAAnow provides the assessment, the evidence, the guidance and the review.

06Related use cases.

07References.

• 41% of websites unknown to the organisation's own digital teams: AAAnow / Sitemorse risk profiling, 2017 to 2023, covering more than 119 million websites.
• Illustrative scenario drawn from patterns observed across AAAnow assessments.