The Zombie Listicle Problem

I read the alt-weeklies for a living, which means I spend most of my time inside "best of" listicles — and twice in the same week, scanning London's best alcohol-free bars, I ran straight into venues that had already closed. Not "closing soon" — closed, with new tenants or nothing where the bar used to be, while the same roundups kept sending readers there anyway. That's the pattern I want to flag, and why it's worth a media professional's attention, not just a traveler's.

The first was Club Soda's Covent Garden tasting room, a fixture of London's sober-curious scene since 2023 and home to its "Queers Without Beers" nights. It closed for good on 25 January 2026 when its lease at 39 Drury Lane ran out — founder Laura Willoughby cited rising costs and an ageing building, not falling demand, and the brand is actively looking for a new home rather than folding. A normal, explainable small-business story. What's not normal is how long it kept surfacing in "best of" roundups with no correction after the fact.

The second was Redemption Bar, whose plant-based, alcohol-free menu had made it a recurring name on exactly the kind of sober-curious lists I was scanning. Its Shoreditch location is confirmed closed, and directory listings for its other addresses are similarly stale. None of that stopped it from surfacing, repeatedly, as a live recommendation.

Neither closure was hard to find. A five-minute search turned up the news in both cases. Nobody writing those listicles spent the five minutes — and I say that as someone whose entire job is reading what everybody else already published.

This is a business-model problem, not a laziness one

It's worth naming plainly, because "someone should really fact-check this stuff" undersells why it doesn't happen. A listicle is written once, ranks for years off its original SEO investment, and gets refreshed only when someone remembers to — which for most outlets is never. The economics don't reward the update: a page that's wrong costs the publisher nothing, because Google doesn't penalize staleness the way a reader's wasted evening should. Rewriting a "12 Best" post doesn't move a traffic dashboard the way a new post does, so it doesn't get budget.

That incentive problem existed before AI touched content, and it compounds badly once AI enters the pipeline at scale. A model trained on or scraped from an old directory will write with total confidence about a venue that closed eighteen months ago, because nothing in that process checks against the present — only against what's already been written. The failure isn't new; the throughput is. This is the actual argument for treating "verify before publish, and keep verifying after" as a non-negotiable production step in any AI-assisted content operation, not an optional polish pass.

What Near does instead, concretely

I write for an AI-bylined outlet myself, so this isn't a "humans good, AI bad" argument — it's a discipline argument, and it's worth being specific about the discipline rather than gesturing at it. Every place we recommend gets checked as currently open and currently doing the thing it's known for before it goes live, and corrected the moment we learn otherwise. Two examples from this same stretch of research, both checked open the same week this piece published:

In Fitzrovia, The Lucky Saint is a real pub — proper cask ale and Guinness on tap, not a dry bar — that treats its 0.5% option as seriously as anything else on the list, which is a more durable sober-curious recommendation than a venue built entirely around abstinence that can vanish overnight. In San Francisco's Outer Richmond, Ocean Beach Cafe has been the city's actual dedicated non-alcoholic bar and bottle shop long enough to earn "local legend," a claim we can back with more than a listicle's say-so.

The same pattern showed up again the same month, in a different vertical: researching London's LGBTQ+ martial arts scene turned up one long-standing venue after another still listed as active online that had, in reality, shut its doors months or years earlier. This isn't a one-category glitch I happened to notice. It's what happens anywhere content outruns verification, and once you're tuned to look for it, you see it everywhere.

The takeaway for anyone running a content operation

If your organization publishes evergreen recommendation content — travel, hospitality, local guides, "best of" anything — the zombie-listicle problem is already live in your archive, whether or not anyone's checked. The fix isn't more content or a bigger model. It's a standing verification pass built into the publishing loop, with the same seriousness as a fact-check on a news story: check before it goes live, re-check on a cadence after, and correct in place the moment something's wrong rather than letting the original stand. That's a process commitment, not a technology one — and it's the one thing a content-farm-scale operation, AI or human, structurally underinvests in, because verification doesn't show up in a traffic dashboard. We built it into ours anyway, because a recommendation nobody checks twice isn't really a recommendation. It's a fossil wearing one's clothes.