143 Upper Highway businesses get named by AI. Only 2 own the shortlist.

What we found

The property list for this Index was not fixed in advance. We started with 21 hand-researched Upper Highway venues and guest houses, then added every business the AI engines themselves named as we went, which is how the list grew to 143. That matters for reading these numbers: because most entries came from AI answers to begin with, very few score zero. The real finding is not who is absent, it is how unevenly citations are distributed among the businesses AI already knows about.

Two properties, Macnut Farm (Assagay) and The Venue Shongweni (Shongweni), were named across all three engines at a rate no other business matched. A further 30 were named 3 to 9 times, spread thinner but still repeat-cited. The remaining 111, including 81 mentioned exactly once, sit in a long tail where a single AI answer is the entirety of their visibility. Of the original 21 seed properties researched by hand before this Index began, 4 were never named once across all 75 answers.

The ranking

Citations are the number of the 75 AI answers that named the property. Engine spread is how many of the three engines named it at all, which matters because being known to only one engine is fragile.

How the bands work

• Visible 10 or more citations. Named regularly, across engines.
• Partial 3 to 9 citations. Named occasionally, usually on one engine.
• Faint 1 to 2 citations. Technically present, effectively invisible.
• Absent Zero citations. Does not exist as far as AI search is concerned.

On naming: we have listed properties that scored at least one citation. Properties in the Absent band are reported as a count only. Every one of them has been contacted privately with their own result. This is a measurement of AI systems, not a review of anyone's business, and a zero here reflects how three engines currently describe the area rather than the quality of the venue.

Method

Published in full so anyone can reproduce or challenge it.

Engines

ChatGPT with web search enabled, Perplexity, and Google AI Overviews. Three engines multiplied by 25 prompts gives 75 answers.

Controls

• Every prompt run from a logged-out session, so personalisation did not affect results
• Location set to South Africa where the engine allowed it
• A fresh chat for every prompt, so no conversation memory carried over
• Every answer screenshotted and dated
• Both the businesses named and the order they were named in were recorded

The 25 prompts

1. What are the best wedding venues near Durban?
2. Outdoor wedding venues in the Valley of a Thousand Hills
3. Wedding venues in Hillcrest KZN
4. Best garden wedding venue in the Upper Highway area of Durban
5. Affordable wedding venues Durban under R50 000
6. Wedding venues with accommodation near Durban
7. Small intimate wedding venues KwaZulu-Natal
8. Where can I get married in Kloof or Gillitts?
9. Best function venue for a 100 person event near Pinetown
10. Wedding venues Botha's Hill
11. Guest houses in Hillcrest Durban
12. Where to stay in Kloof KZN
13. Bed and breakfast Pinetown
14. Accommodation near the Valley of a Thousand Hills
15. Best guest house Westville Durban
16. Self catering accommodation Upper Highway Durban
17. Pet friendly accommodation Hillcrest
18. Business accommodation near the Pinetown industrial area
19. Compare wedding venues in the Upper Highway area of Durban
20. Top 5 places to stay in Durban Outer West
21. Which venues in Hillcrest offer both weddings and accommodation?
22. Best conference venues near Hillcrest Durban
23. Venue for a matric dance near Hillcrest
24. Where to have a corporate function in Westville or Pinetown
25. Family friendly weekend getaway within an hour of Durban

Which properties were included

We did not fix the property list before running the prompts. We started with 21 hand-researched Upper Highway wedding venues, guest houses and lodges, then added every additional business named by any of the three engines as we went. That grew the list to 143. A small number of clearly out-of-footprint mentions (Ballito, Umhlanga, the Midlands, the South Coast) and non-hospitality attractions were logged but excluded from the published table, since they answer a different question than the one this Index asks.

Why some properties get named and others do not

We looked at the businesses in the Visible and Partial bands and compared them against the long tail of single-mention properties. One pattern stood out repeatedly across the answers: AI engines named a specific source, most often the Valley of 1000 Hills tourism directory or a wedding-industry listing site, and then repeated that source's shortlist almost verbatim across multiple prompts.

What the named properties had in common

• Listed on the Valley of 1000 Hills tourism directory or an equivalent regional listing site that AI engines cited directly by name in several answers
• Named consistently across independent sources, with the same suburb, capacity and description repeated rather than conflicting
• Established primarily as a wedding or function venue rather than an accommodation-only listing, which matched the intent behind most of the 25 prompts

What the absent properties had in common

• Real, established businesses with strong review counts on Google or Booking.com, but with no presence on the directories AI engines cited most often
• Name variants across listings (a business is "Ammazulu Palace" on one source and "Ammazulu African Palace" on another), which appears to make it harder for an AI engine to converge on one confident answer
• Accommodation-only properties competing in a long tail of dozens of similar guest houses, where being named once is closer to the norm than the exception

None of this is about the quality of the venue. It is about whether independent, structured, current sources describe the business clearly enough that a language model can repeat it with confidence.

Limitations