How to tell whether a holiday listing still exists

Everything below comes from work we did for our own guidebooks. We built venue lists from mapped open data, then opened and read every candidate's own website one at a time, recording the date. Across six destinations that came to 1,621 sites.

The failure rates were not what we expected, and they varied enormously by place.

What we measured

| Destination | Candidates checked | Dead or repurposed | |---|---|---| | Tuscany | 27 | 6 per cent | | Amalfi Coast | 20 | 25 per cent | | Marrakech and the Atlas | 107 | 28 per cent | | Sri Lanka | 356 | 32 per cent | | Bali | 905 | 41 per cent |

Sites that merely blocked an automated request were excluded from those counts, because a bot defence is not a dead business and counting it as one would flatter the method.

Then we did the Mani, in southern Greece, and the measurement stopped making sense. The peninsula has 206 mapped eating and drinking places and 17 websites between them. The question there is not what proportion had died. It is that 92 per cent never published anything in the first place.

The three failure modes, in order of how much damage they do

**Dead.** The domain does not resolve. Annoying, and harmless, because it announces itself.

**Repurposed.** Somebody else owns the domain now, and this is not occasional. In Bali we found **13 former tourism domains serving online gambling sites**: restaurants, hotels and hostels whose addresses now return Indonesian slot and sports-betting affiliates, plus one Dutch casino and one Czech one. In Sri Lanka, six former hotel and restaurant domains had become gambling or celebrity-scandal sites. In Morocco, one restaurant's own domain was serving an Indonesian gambling site.

Three destinations on three continents, and repeatedly the same kind of operator. This is a business model rather than a coincidence: an expired tourism domain carries aged inbound links from travel sites, guides and directories, and that link equity is worth buying. Online gambling is illegal in Indonesia, which is presumably why the sites sit on foreign-registered domains that used to belong to somebody else's restaurant.

**Compromised.** This is the dangerous one. In the Mani we found a bakery, trading on a claim of continuous operation since 1890, whose site returns a clean page with its own branding intact: the right name, the right products, the right photographs. Sitting in the same homepage HTML were 34 instances of the Greek word for casino, in gambling affiliate copy, with an outbound link to an operator.

The site was not abandoned and the domain was not sold. It was compromised, and the sentences that look like the business talking are interleaved with sentences that are not. **A page that looks normal is not evidence that it is.**

The near-miss, which is worse than a dead link

While researching the same region we tried to resolve venues by guessing their domains. Greek business names are generic and they collide nationally. The obvious domains returned an Italian pharmaceutical laboratory, a Croatian architecture portal, a hotel on a different island and an investment company. Every one of those is a real business with a real, functioning website.

There was also a cafe whose third-party listing published a website that does not resolve, while a very similarly named live domain belonged to a completely different company.

**The reader who guesses lands on somebody else's business and has no way of knowing.**

Five checks that catch most of it

1. **Find the site from the business, not the business from the site.** A listing's link is a lead. Confirm it against a phone call, a map entry, or a social account the business obviously runs. 2. **Read a page that is not the homepage.** Compromised sites are often injected on some pages and not others. The menu, the contact page and the news page are where it shows. 3. **Check the copyright year and whether anything has been updated.** Not conclusive, but a 2019 copyright on a hotel is worth a phone call. 4. **Distrust third-party opening hours completely.** Across everything we tested, this was the least reliable category of information. During Tuscany research, a mapped listing showed a Chianti restaurant open through the winter when its own site says it shuts from December to February. 5. **Telephone.** Especially anywhere rural. In the Mani it is not the fallback, it is the entire system, because for most businesses there is nothing else.

The uncomfortable conclusion

The failure rate is a property of the destination, not of your carefulness, and it does not track how developed a place is. Bali is heavily mapped, heavily commercial and comprehensively online, and it had the worst rate we have measured.

What it tracks is business size. In Bali the failure rate ran at 61 per cent for hostels, 50 for cafes, 49 for guest houses, 42 for restaurants and 36 for hotels. By region it barely moved at all. Small independent businesses churn, their domains lapse, and somebody is waiting to buy them.

So the practical rule is the opposite of the intuitive one. **The smaller and more interesting the place you are trying to book, the less you should trust the link.**