When Yesterday’s Coverage Becomes Today’s Answer

This week Bo House, one of the flagship Mediterranean brands of Spotlight client Eterniti, landed a glowing review in The Times. The article will remain online long after some of the details it records have changed. So what happens when AI finds it years from now? Alexandra Harvey on the problem of outdated content in AI search.

The one thing rarer in St Tropez than a parking spot in the old town: access.

Julia Llewellyn Smith, The Times

So wrote Julia Llewellyn Smith in The Times a few days ago, in the kind of piece PR teams dream about. The subject was Bo House, the Mediterranean villa and property brand within Eterniti, whose St Tropez director, Augustin Hacot, Julia described as being “on cheek-kissing terms with everyone you need to know”. She went on to explain how Hacot went to school with the maître d’ at Le Club 55 and how the numbers for Les Caves du Roy were in his Filofax long before mobile phones were invented.

It’s funny, specific and wonderfully detailed. From a PR point of view, a triumph.

So here’s an odd question: what happens when some of it is no longer true?

The Filofax anecdote will stand the test of time. But the same article records that Bo House has more than 300 villas and apartments and that a single villa can cost about £135,000 a week in high season.

All true today. But what about in 2030?

Until recently, the influence of a feature like this would generally diminish with time. It would slip down search results and deeper into the publication’s archive. It would still have value, but its moment of greatest visibility had passed.

But AI has changed that.

Ask an AI assistant for a hotel or villa recommendation and it can draw on years of material from across the internet, piecing together reviews, hotel websites, travel guides and news stories to construct an answer. Suddenly, old coverage has an afterlife.

And sometimes it’s telling yesterday’s story. Hotels are refurbished. Chefs leave. Restaurants close. Prices rise. Brands change hands, change names or completely reposition themselves. The journalism that recorded them at a particular moment, quite rightly, remains.

What’s changed is who, or what, is reading it, and what they do with the information they find.

When the past becomes evidence

One of the more intriguing findings from Invisible or Influential?, our research with Make Lemonade into how 30 leading travel and luxury publications are adapting to AI, was just how messy the past can become.

Right now, Bo House offers a useful example of what happens when the story is consistent. The Times piece isn’t the only recent article telling its story. A recent OutThere review recorded many of the same facts: a portfolio of more than 300 villas and apartments and, importantly, the local relationships that distinguish it from its competitors.

For AI retrieval, that consistency can be useful. Two independent publications are reinforcing the same current picture of the brand.

Fast-forward a few years, however, and the picture can become considerably messier. An old review says one thing, a newer feature says another, while the website says something else entirely. Conflicting stories, duplicated articles and outdated recommendations can all become part of the evidence AI uses to decide what’s true now.

Archive is no longer passive infrastructure. It is an active visibility signal.

If freshness matters, why does AI still get things wrong?

Here’s the apparent contradiction. Research by Ahrefs looked at around 17 million URLs cited by AI assistants and found that cited pages averaged approximately 1,064 days old, compared with 1,432 days for pages ranking in Google’s organic top ten. In other words, the pages being cited by AI were around 25.7 per cent fresher.

Our own research pointed to the other side of the same equation. It isn’t only what sits in the archive that matters, but what has been published since: recent, aligned coverage can help establish whether an older recommendation, description or editorial position still holds true.

That doesn’t prove that making something newer automatically makes AI more likely to cite it. But it does suggest that recency matters, particularly when the question itself depends on what is true now.

And few questions depend more on the present than travel recommendations. Nobody asking where to stay in St Tropez wants the answer from four years ago.

So why does AI still confidently serve up old names, departed chefs and restaurants that closed years ago?

Because AI can arrive at an answer in different ways.

Some of what a large language model knows comes from material absorbed during training. That historical information doesn’t magically disappear because something newer has subsequently been published.

At other times, AI systems retrieve information from the web to help answer a question. Here, newer and well-maintained information has a much better opportunity to establish what is true today.

Which leaves brands with an awkward reality: you may have moved on, while parts of the internet haven’t.

You can’t rewrite the past

There is another complication for PR teams. Most of the information about your brand doesn’t belong to you.

If a newspaper review from 2021 contains an old room count, you can’t log in and change it. Nor, in most cases, should you expect the publication to retrospectively rewrite a perfectly accurate article because the facts later changed.

This isn’t an argument for constantly ‘correcting’ old journalism. Nor is outdated archive material evidence that publishers are doing something wrong. But publishers are beginning to grapple with the problem too.

Uwern Jong at OutThere knows the problem first-hand. The publication has been going through its substantial back catalogue, reviewing, reworking and, where necessary, archiving legacy stories, around 700 articles in all. “It’s heartbreaking as a writer,” Jong says, “but archiving is key.”

The problem wasn’t simply that some of those old stories were no longer useful to readers. Jong told us they were also affecting OutThere’s AI capabilities, which were pulling up outdated information, including hotels that no longer existed and travel tips that were no longer relevant.

That doesn’t mean old journalism is inherently a liability. Quite the opposite: years of trusted coverage can give an established publication an authority a new website simply cannot manufacture overnight. But the OutThere experience illustrates the new challenge for publishers: an archive can no longer simply be somewhere journalism goes when its first life is over. Increasingly, somebody has to know what’s still in there.

It’s no small undertaking. Archive management is a huge and rather unglamorous job. Responsibility often sits somewhere between editorial, digital product, SEO and commercial teams, which can mean it belongs completely to nobody. For a magazine publishing hundreds or thousands of stories a year, continually auditing everything it has ever published would be an enormous task.

For brands, therefore, the more useful question isn’t how do we get rid of the past?

It’s how do we make the present easier to find?

The best correction may not be a correction

This is where PR strategy gets interesting.

For years, communications teams have thought primarily about securing coverage. Where did the story appear? Who wrote it? How many people saw it? What was its reach?

AI adds another question: how long will the information inside it remain useful?

Some stories age beautifully. A profile of a hotel founder, an interview with an architect or a feature about a property’s history may become more authoritative with time.

Others have a much shorter shelf life. Hotel rankings, restaurant recommendations, prices, openings, ownership, leadership and practical destination advice can date remarkably quickly.

And because brands cannot control most earned coverage once it has been published, chasing every outdated reference around the internet isn’t a realistic strategy.

Our research points towards a more practical answer:

Consistency displaces stale facts more reliably than correction does.

If an old article contains information you cannot change, make the current story clearer, more authoritative and easier to find elsewhere.

That starts with the part of the internet you can control. Your hotel website, property pages, press room, fact sheets and news pages should contain unambiguous, current information about names, dates, room counts, ownership, leadership and positioning.

But owned content can’t do all the work. Current earned coverage matters too, not simply because somebody might read the article, but because it adds independent, authoritative evidence about what the brand is today.

Publishing rhythm matters for the same reason. Recent, consistent information helps establish whether an older recommendation or description still holds true. One new article won’t necessarily displace a claim repeated across ten years of archive. A coherent body of current information has a much better chance.

None of this makes yesterday’s coverage worthless. Far from it. Authority accumulates.

But placement and permanence are no longer quite the same thing. A feature can be enormously valuable on the day it appears while some of the facts inside it become less useful over time.

PR has always been about influencing what people know about a brand.

The new challenge is making sure yesterday’s truth doesn’t become tomorrow’s answer.

FAQs

Does AI search prefer newer content?

For live retrieval, generally yes. Ahrefs’ analysis of around 17 million AI citations found cited pages were roughly 25.7 per cent fresher than pages ranking in Google’s organic top ten. The effect matters most for queries where the answer depends on current facts, which includes most hotel and destination recommendations.

If AI prefers fresh content, why does it still repeat outdated information about my hotel?

Because two mechanisms are at work. Some information comes from what a model already holds from training, which has no recency filter, rather than from a live page fetch. Fresh content competes well in retrieval but does not automatically overwrite what a system already holds.

Is coverage in a monthly or quarterly title still worth having?

Yes. It does what it has always done for brand authority and credibility. It is simply less likely to be reinforced by surrounding recent coverage, which AI systems can use to validate that a position still holds. The two outcomes are worth planning separately.

Does a paywall prevent AI citation?

Not necessarily. Our research found that what matters is not simply whether content is paid for, but how AI platforms are able to access it. Licensing agreements, training-period accessibility and registration layers can all produce different outcomes, and paywalled journalism can still surface prominently.

How often should a hotel update its own pages?

Often enough that the core facts are never wrong: names, dates, room counts, leadership and positioning. What matters is a substantive update. Simply changing the visible publication date without changing the content doesn’t help.