01What happens.
19% of PDFs on organisational websites are duplicates, and 37% become untracked once they leave the producer's site, through circulation in catalogues, partner sites and reports. Duplicates send inconsistent signals about which version is current, and correcting one copy leaves the others in place.
Some copies are found and some are not, and the ones not found keep answering.
02Why it matters.
The head of digital content owns the asset and republished it correctly. What the team cannot see is how many copies of the old one exist, where they are, and which of them AI is reading. Each correction the team makes is measured against an answer that does not move.
The management burden is a team repeatedly chasing copies it cannot list while the old instruction remains retrievable. Preventing the same problem requires a distribution record and document-control rules shared by the teams that publish. The AI Accelerator's internal education addresses those rules before the next leaflet is issued.
03The case.
A consumer brand (we do not mention names / we ammonise use cases) finds an error in a product leaflet, a dosage instruction on a household product, and corrects it. The new leaflet goes onto the product page and into the document store, and the old file is overwritten there. The correction is done properly by the standards the brand applies.
The old asset has been online for four years, in that time it has been copied into a regional site, attached to a promotional microsite and uploaded twice more to the document store under different names. It has been placed in multi retailers' catalogues, hosted by a comparison site, and downloaded by customers who forward it to each other. Most of the copies are identical to the original.
A customer asks AI how much of the product to use. Several old copies remain accessible across brand, retailer, comparison and customer-held locations, while the corrected version is new. Their age, links and duplication make the old instruction appear well established, so the customer receives the instruction the brand corrected.
The content team finds a retailer copy and asks for it to be replaced, and the answer does not change. The team finds the regional copy and replaces it, and the answer does not change. Each round takes weeks, each removes one copy, and the copies the team has not found go on supplying the answer.
The correction was made properly at source, but nobody had recorded the wider estate. The old leaflet now existed in multiple locations, some controlled by the brand and others outside its control, which the content team could not fully list.
04Exposure.
The brand is held to the instruction in circulation under its name. Where the instruction concerns the safe use of a product, the old leaflet is a safety exposure for as long as any copy of it answers. Correction at source does not discharge it, because the copies were never at source.
The cost repeats: content team time spent finding copies, partner correspondence to have each replaced, and checks after each round while the old instruction remains retrievable. Without an inventory and distribution controls, later corrections can produce the same chase.
05What AI Readiness changes.
The team was correcting the copies it could see and being judged on the ones it could not. The outside-in reading AAAnow provides lists the copies: each duplicate on the brand's own properties, each untracked copy on retailer, partner and comparison sites, and which of them AI is drawing on for the question customers ask. AAAnow evidences that list and the answer AI returns.
A single source, held centrally against a governance model that can is monitored – would have addressed this issue….
With the list in hand, the brand retires the duplicates it controls, takes the evidence to each partner that holds a copy, publishes the corrected leaflet in a machine-readable form, and checks which old copies still influence AI responses. Document-control rules are shared across publishing teams, so later corrections begin with a record of where controlled copies were distributed.
06Related use cases.
Drafts uploaded to the document store were served as the company's final word
The repository in which duplicates and drafts accumulate.
20 years of documents were still answering for the company as if each were current
The legacy estate(s) that holds the copies.
The impact of out-of-date safety information: who is liable?
A retailer's copy outranking the brand's.
07References.
- 19% of PDFs on organisational websites are duplicates; 37% become untracked once they leave the producer's website: AAAnow / Sitemorse analysis.
Next step
Talk it through, in your terms.
The briefing sets this case in your organisation’s terms, and the discussion follows your questions.
Illustrative scenario, on AAAnow figures. No company is identified.