WebsiteSpa Research · November 2026
State of AI readiness: UAE, November 2026
We put 19 of the UAE’s best-known websites through the same 100-point scan anyone can run in twelve seconds. The country uses AI more than any other on earth. Its websites are not ready for it.
73.3%
of the UAE’s working-age population used a generative AI product in June 2026
4 of 19
top UAE sites are fully ready: they publish an llms.txt, ship structured data and block none of the seven primary AI crawlers
Ready: ziina.com, nymcard.com, eyewa.com, kitopi.com
Key findings
The numbers that decide whether AI can use a UAE website
50.6 /100
Average score
The median is 52. Half the list sits at or below it, and the gap between the best site (ziina.com, 83) and the worst (dubizzle.com, 7) is the story of this edition.
6/19
publish an llms.txt
13 leave the map out entirely, so an assistant has to guess which pages matter and which are noise.
8/19
publish structured data
11 publish no JSON-LD at all, so nothing in their HTML says who the business is, what it sells or where it is.
4/19
show AI almost nothing without JavaScript
Most AI crawlers fetch the HTML and never run JavaScript, so a page assembled by scripts is a blank page to them. These serve under 200 characters of readable text: kitopi.com, aldar.com, du.ae, dubizzle.com.
3/19
block a primary AI crawler
du.ae turns away 6 of the 7 assistants people ask for recommendations. A blocking rule is permission for AI to stay silent about you.
The ranking
Every site, highest score first
Each score is the headline number from that site’s own report — open “View report” to read the checks behind it. Filter by sector to compare like with like.
Showing 19 of 19 sites, highest score first
- 01
ziina.comFintech
83 out of 100. Ready for AI.- llms.txtpublished
- schemapresent
- crawlersall allowed
- 02
nymcard.comFintech
78 out of 100. Workable, with gaps.- llms.txtpublished
- schemapresent
- crawlersall allowed
- 03
eyewa.comE-commerce
75 out of 100. Workable, with gaps.- llms.txtpublished
- schemapresent
- crawlersall allowed
- 04
emaar.comReal estate
71 out of 100. Workable, with gaps.- llms.txtpublished
- schemapresent
- crawlers3 blocked
- 05
careem.comFood & delivery
61 out of 100. Workable, with gaps.- llms.txtmissing
- schemapresent
- crawlersall allowed
- 06
sarwa.coFintech
59 out of 100. Workable, with gaps.- llms.txtpublished
- schemamissing
- crawlersall allowed
- 07
cafu.comFood & delivery
57 out of 100. Workable, with gaps.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 08
kitopi.comFood & deliveryJS onlyWithout JavaScript, AI crawlers see almost nothing here. This homepage serves 136 characters of readable text in its raw HTML.
57 out of 100. Workable, with gaps.- llms.txtpublished
- schemapresent
- crawlersall allowed
- 09
getstake.comReal estate
52 out of 100. Workable, with gaps.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 10
propertyfinder.aeReal estate
52 out of 100. Workable, with gaps.- llms.txtmissing
- schemapresent
- crawlersall allowed
- 11
zbooni.comE-commerce
50 out of 100. Workable, with gaps.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 12
bayzat.comHR & services
48 out of 100. Hard for AI to use.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 13
huspy.comReal estate
46 out of 100. Hard for AI to use.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 14
wio.ioFintech
44 out of 100. Hard for AI to use.- llms.txtmissing
- schemamissing
- crawlers2 blocked
- 15
talabat.comFood & delivery
38 out of 100. Hard for AI to use.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 16
tabby.aiFintech
36 out of 100. Hard for AI to use.- llms.txtmissing
- schemapresent
- crawlersall allowed
- 17
aldar.comReal estateJS onlyWithout JavaScript, AI crawlers see almost nothing here. This homepage serves 17 characters of readable text in its raw HTML.
28 out of 100. Hard for AI to use.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 18
du.aeTelecomJS onlyWithout JavaScript, AI crawlers see almost nothing here. This homepage serves 13 characters of readable text in its raw HTML.
19 out of 100. Hard for AI to use.- llms.txtmissing
- schemamissing
- crawlers6 blocked
- 19
dubizzle.comE-commerceJS onlyWithout JavaScript, AI crawlers see almost nothing here. This homepage serves 6 characters of readable text in its raw HTML.
7 out of 100. Hard for AI to use.- llms.txtmissing
- schemamissing
- crawlersall allowed
- 80–100 ready
- 50–79 workable
- 0–49 hard for AI to use
By sector
Fintech leads. Nobody is comfortable.
The average score of each sector, out of 100. The lowest average belongs to Telecom, which has a single site: read a sector that small as a data point, not a rule.
Bars use the same bands as the table: 80+ green, 50–79 blue, under 50 coral. 4 of 6 sector averages sit under 50. The band describes the sector’s average, not its best site.
What the leaders do
The top three share three habits the rest skip
Top three: ziina.com, nymcard.com, eyewa.com. Each card below is a real check from those reports, with how many of the other 16 sites pass the same check.
3/3 vs 1/16
llms.txt links resolve
They do not just publish an llms.txt — every link in it resolves to a page that exists. An llms.txt full of dead links teaches an assistant that the map is wrong.
3/3 vs 2/16
JSON-LD entity identity
Organization plus WebSite JSON-LD, with a name, a URL and the profiles that corroborate it. It is the difference between a page and a business an assistant can name.
2/3 vs 0/16
Organization schema completeness
The Organization block is filled in: logo, contact point, address. When only name and url are present, an assistant has nothing to quote when someone asks how to reach you.
Fix your site in an afternoon
Three changes move almost any site up the board
These are the three checks that separate the top three from everyone else. None of them needs a developer week, and all three are free to do.
Publish an llms.txt
A markdown file at your domain root: H1, a one-line summary, then linked sections for the pages that matter. It is the map an assistant reads first.
Check your llms.txtShip Organization schema
name, url, logo, contactPoint and sameAs on every page. It is how an assistant decides that two pages are the same business.
How to write itLet the crawlers in
GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended, allowed in robots.txt. A blocked crawler is an assistant that can never recommend you.
Read the rules
Your turn
Put your own site through the same scan
Twelve seconds, no account and no card. You get the six pillars, every check with its evidence, and the free Agentify kit: the llms.txt, crawler rules, schema and FAQ block this edition is scoring you on.
Track it weekly
PRO, $10 a month
The same questions your buyers ask, put to ChatGPT, Perplexity, Gemini and Claude every week, with who they name instead of you. Not on sale yet — the waitlist is how the beta opens.
MethodologyHow these scores were produced
One score, six pillars
Every site gets one 100-point score. It is built from the six pillars below, weighted the way the scanner weights them for every site it scans — the same engine, the same weights, no edition-specific rules.
- Crawler access25
- llms.txt10
- Entity & structured data20
- Answer readiness18
- Brand clarity15
- Freshness & trust12
- Total100
The report shows no pillar subtotals. The headline score is the number the methodology defines and /scan/<domain> prints, and the pillar points underneath it are on a different scale — adding them up is not the same sum. Every score here links to the report it came from, so nothing has to be taken on trust.
What was scanned
Public pages only: the homepage, plus the pages the site itself links as important. No login, no paywall, nothing private. The scan reads robots.txt, llms.txt, the raw HTML and a sample of pages for each site.
Letters are the scanner’s own bands for that number: 90+ A, 80 A−, 72 B+, 64 B, 56 B−, 48 C+, 40 C, 30 C−, 20 D, below that F.
“Fully ready” in the hero is not a score band: it means all three technical essentials are in place — an llms.txt, structured data, and no primary AI crawler blocked.
Dates, changes, contact
Every site in the table was scanned on 14 Nov 2026; each row links to its report, which carries its own snapshot time. A site that changes gets a new report the moment it is rescanned, and the next edition is measured from scratch.
Spotted something wrong, or want a site added or rescanned? Email hello@websitespa.net with the domain and we will run it again and correct the page.
Selection: 19 sites across 6 sectors, chosen for reach and brand recognition in the Emirates. This is a sample, not a census of every UAE domain.
Couldn’t be scanned
6 sites on the original list have no score in this edition. They are listed, not guessed: a missing number is a finding, and a made-up one would poison the rest of the table.
- bayut.comThe homepage answered 503 to a plain request, so the scanner never received a page to score. Listing it with a number would measure the wall instead of the site.
- instashop.comThe homepage answered 403 to a plain request, so the scanner never received a page to score. Listing it with a number would measure the wall instead of the site.
- emirates.comRetried once on 14 November 2026 through the site's own scanner: could not reach emirates.com. A normal browser request is answered with a 403 access-restricted page, so an automated scan cannot get past it.
- noon.comRetried once on 14 November 2026 through the site's own scanner: the scan was still running after 150 seconds. The homepage answers with a redirect to itself, so a scanner never lands on a page it can read.
- majidalfuttaim.comRetried once on 14 November 2026 through the site's own scanner: the scan was still running after 150 seconds. The site does answer a browser request, so this is a failure inside the scan run, not a wall.
- pureharvest.aeRetried once on 14 November 2026 through the site's own scanner: could not reach pureharvest.ae. Its TLS certificate expired in July 2025, so a scanner that validates certificates stops at the handshake.
Next edition: the same 19 sites, rescanned, plus the sectors this sample could not cover.
Published 14 Nov 2026