01What happens.
The employer AI describes to candidates is assembled from the digital landscape, not from the current careers site. Old careers pages, expired adverts on job boards, past benefits statements and third-party reviews persist, each carrying the company's name, and a machine weighs them as readily as the current page (especially if they are not dated).
02Why it matters.
The head of talent rewrote the careers site and the offer it describes. Some candidates ask AI before visiting the careers site. In a 2026 survey of talent acquisition leaders, 73% were concerned that candidates receive outdated or inaccurate information about their organisation from AI, and 33% were confident their employer brand is accurately represented in AI search (Built In).
An applicant who accepts on a benefit that no longer exists arrives with a grievance, and an applicant who declines on a description that no longer applies never arrives at all.
03The case.
The consumer brand overhauls its employment offer: the benefits package changes, the hybrid working policy is rewritten, dule job families are closed and 1 is created. The careers site is rebuilt to describe the new offer, and the recruitment team briefs its agencies.
The old offer sits there in the places the rebuild did not reach. An earlier careers microsite, built for a graduate campaign, still describes the previous benefits and the roles that were closed, and expired adverts on job boards carry the old package in their standard text. A benefits explainer PDF sits in the document store under its old name. Third-party review sites carry descriptions written by employees under the old arrangements.
A candidate asks AI what the company offers and what it is like to work there. The answer draws on the microsite, the expired adverts and the PDF, because between them they state the old offer many times, while the new careers site states the new offer once. The candidate is told about a pension contribution the company reduced, a role family that was closed, and a working pattern that know longer applies.
Some candidates submit applications through stale listings for roles that no longer exist and are told so at screening, after the company has paid to process the application. One accepts an offer and raises the missing benefit in the first week. Others compare the AI description with a competitor's and go elsewhere, and the company never learns why, because a candidate who does not apply does not explain.
The recruitment team rebuilt the careers site correctly. The rebuild did not include an inventory of where the old offer had been published, because the previous offer had been published over years by several teams, and no such inventory existed.
04Exposure.
The company is accountable for employment information it publishes or authorises. Third-party pages and AI summaries can still create expectations that the company has to address, but they are not automatically the company's own statements. A candidate who accepts a role expecting a benefit that no longer exists raises a dispute the company has to resolve, while a candidate who declines on an outdated description creates a loss the company cannot see.
Recruitment spend is diluted by a description the company has already replaced, and the employer brand the company invested in is outvoted by its own past. The company can count the candidates who raise the problem; it cannot count the ones who read the old offer and went elsewhere without a word..
05What AI Readiness changes.
The rebuild replaced the careers site and left the old offer where the team could not see it. Read from outside, as a candidate's AI reads it, the employer the company presents is assembled from the microsite, the job boards, the document store and the review sites. We offer evidence which of those AI is drawing on and what it tells a candidate. On that the company retires the old careers material it controls, asks job boards and agencies to retire the expired adverts, publishes the current offer in a machine-readable form and checks what AI tells candidates after those changes.
06Related use cases.
Customers were shown a brand the company had retired, from guidelines it had replaced
The same weight-of-record mechanism applied to the brand.
Investors were told the former chair still ran the board
The same mechanism applied to leadership.
20 years of documents were still answering for the company as if each were current
The legacy estate that holds the old careers material.
07References.
- Built In, 2026: 73% of talent acquisition leaders concerned that candidates receive outdated or inaccurate information from AI; 33% confident their employer brand is accurately represented in AI search. https://builtin.com/employer-branding/ai-employer-brand-reportbuiltin.comWe do not enable external links - please copy and paste, you are then certain as to the link being opened
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, drawn from industry research. No company is identified.