The GCC Business Website Index 2026

Everybody in this industry has an opinion about the state of business websites in the Gulf, and nobody publishes a number. So we measured. The businesses came from OpenStreetMap, the measurements came from the same checker anyone can run at mabni.co/tools/website-check, and the whole dataset is at the bottom of this page for anyone who wants to check the arithmetic.

The single most striking result is not about design or speed. It is that more than a quarter of these businesses are advertising an address that goes nowhere.

More than a quarter of these websites are broken

Never answered

288 (19.3%)

Answered with an error

127 (8.5%)

Served a working page

1,030 (69.0%)

Blocked our crawler in robots.txt

16 (1.1%)

Before counting a site as dead we fetched it a second time with a normal browser user agent and a thirty-second timeout, then once more after that. 29 sites recovered on the second attempt and were moved back into the working column. The 415 that remained had three chances and took none of them.

A dead website is worse than no website. The address is still on the business card, still in the map listing, still in the WhatsApp message a customer forwards to a friend. Every one of those is now a link to nothing, and it reads as a business that has closed.

A further 88 businesses listed a Facebook or Instagram page where a website should be. Those are excluded from every figure on this page, because a social profile is a different thing with different rules, but they are worth knowing about: they are businesses that decided a website was not worth having.

They told a volunteer map project their hours, and not their own website

This is the finding we did not expect. 672 of the businesses we measured had published their opening hours to OpenStreetMap, which means somebody sat down and typed them into a free, volunteer-run map. Of those 672, only 18.5% state the same hours on their own website in a way a machine can read.

Marked-up opening hours are how "open now" works in Google Search and in Maps. Hours typed into a footer as ordinary text are not that. They are readable by a person who has already found you, which is the one audience that no longer needs them.

Across the whole measured sample, 15.9% publish machine-readable hours. For restaurants specifically it is 14.1%, which is the figure worth sitting with: the businesses whose customers most need to know whether they are open right now are the least likely to say so in a form anything can read.

What the twelve checks found

Each site was scored on twelve measurements, all taken from a single fetch of the home page plus its robots.txt and sitemap. These are the pass rates across the 1,030 sites that served a working page.

Served over HTTPS

96.3%

Has a mobile viewport tag

94.3%

AI crawlers allowed

86.3%

Text readable without JavaScript

70.5%

Server answered under 600ms

67.4%

Title of a sensible length

65.1%

Has a tappable phone or WhatsApp link

54.9%

Every image described

55.4%

Structured data identifying the business

49.5%

Description of a sensible length

44.0%

Publishes a sitemap

26.7%

Machine-readable opening hours

15.9%

The basics are largely solved

HTTPS at 96.3% and a mobile viewport at 94.3% are the two things the industry spent a decade arguing about, and they are now close to universal. Nobody needs another blog post about either.

Almost nobody publishes a sitemap

73.3% have no sitemap at /sitemap.xml. It is a file listing your pages, most content management systems generate one automatically, and without it a search engine finds your pages only by following links to them. This is the cheapest unfixed problem in the dataset.

Two in five are invisible as a business

40.5% carry no structured data at all, and a further 10% carry some that never says what kind of business this is. Structured data is how a search engine or an AI assistant knows it is looking at a restaurant in Jeddah rather than a page containing the word restaurant. Without it, everything about the business has to be inferred from prose.

One in six hands a crawler an empty page

16.1% had almost no readable text in the HTML, which means the words are assembled in the browser after the page loads. A person sees a finished page. A search engine and an AI assistant receive a shell. This is the one problem on the list that is usually decided by whatever built the site rather than by anything the owner did.

Nearly half cannot be contacted in one tap

45.1% have no phone number that dials and no WhatsApp link that opens a chat. A number typed as plain text is neither. On a phone, that is the difference between an enquiry and a customer who copies eleven digits by hand, or does not.

Almost nobody blocks the AI crawlers

Only 2.9% block GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot or Google-Extended, and a further 10.8% publish no robots.txt at all. We expected this number to be much higher, because plenty of publishers blocked these crawlers in 2023 and never revisited it. Gulf businesses mostly never blocked them in the first place. The barrier to being quoted by an AI assistant here is not permission. It is having something readable to quote.

Restaurants are the worst of it

Six trades had enough sites to report separately. Restaurants and cafés come last on almost every measure, which is a problem because they are also the businesses whose customers are most often standing outside on a phone.

Clinics and dentists

median 84 · 24.3% broken

Offices and agencies

median 79 · 22.8% broken

Shops

median 77 · 25.5% broken

Motor trade

median 77 · 28.8% broken

Restaurants

median 74 · 34.4% broken

Cafés

median 70 · 29.4% broken

34.4% of restaurant websites in the sample are broken, the highest of any trade. Of the ones that work, 57.9% have no tappable phone number and 25.8% hand a crawler an empty page. Cafés are worse again on contact, at 68.5%.

Clinics and dentists are the clear outlier in the other direction, with a median of 84 and the lowest rate of missing structured data at 26.8%. They are also, of the six, the trade most likely to be spending real money on a website.

The differences between cities are small

Dubai — 943 sites

26.3% broken · median 78

Abu Dhabi — 205 sites

30.2% broken · median 75

Jeddah — 152 sites

27.6% broken · median 76

Riyadh — 182 sites

32.4% broken · median 74

Riyadh has the highest proportion of broken sites and Dubai the lowest, but the spread is six points across four cities and the medians sit within four points of each other. Whatever is going wrong here is not local to one market.

Sharjah was collected and measured, but almost every Sharjah business in the sample shares a domain with a Dubai branch, so after deduplicating by hostname too few unique sites remained to report it on its own.

If you own one of these businesses

Four of these are settings rather than rebuilds, and you or whoever built your site can do them this week.

1. Open your own website address on a phone, on mobile data, right now. A quarter of the businesses in this sample would have found out something they did not know.
2. Publish a sitemap. Most platforms have a switch for it. 73.3% of this sample has not found the switch.
3. Mark up your opening hours, so search and maps can say whether you are open. 84.1% do not.
4. Make your phone number dial and add a WhatsApp link. 45.1% have neither.

The fifth, text that is not readable without JavaScript, is not a setting. If your site is one of the 16.1%, it was decided by whatever built it, and fixing it means rebuilding the pages rather than changing an option.

You can measure your own site against every figure on this page with the free check. It runs the identical code. No sign-up and no email address.

The data

Every aggregate on this page can be recomputed from the file below. Business names and URLs are deliberately left out: nothing is lost to a reader checking the arithmetic, and no named restaurant is held up as an example of a bad website by a company that sells websites.

Download the full dataset (CSV, 1,492 rows)Download the aggregates (JSON)

Use it for anything, including commercially. An attribution and a link back is appreciated but not required. The business locations underneath it come from OpenStreetMap contributors, licensed ODbL.

How this was measured

Businesses were taken from OpenStreetMap through the Overpass API on 29 August 2026, from bounding boxes covering the built-up areas of Dubai, Abu Dhabi, Sharjah, Riyadh and Jeddah. We kept every shop, office, craft business, restaurant, café and clinic that had both a name and a website tag.

Embassies, ministries, government offices, political parties, religious organisations, schools, malls and department stores were removed. They have websites built under conditions that nothing on this page applies to, and leaving them in would have flattered the result.

Results were then deduplicated by hostname, so a chain with nine branches counts once. That took 2,374 listings down to 1,492 unique domains.

Each domain was fetched once, at the home page, along with its robots.txt and its sitemap. robots.txt was honoured: 16 sites disallowed our crawler and were recorded as such rather than measured anyway. Requests identified themselves as MabniResearch with a contact address, ran eight at a time, and paused between calls.

Anything that failed was retried as a browser would retry it, with a familiar user agent, a thirty-second timeout and one further attempt. 29 sites recovered and were counted as working. The measurements themselves come from the same code that runs the public check at mabni.co/tools/website-check, so any figure here can be reproduced against any single site.

What this does not tell you

OpenStreetMap is not a business register. This is a sample of businesses that somebody has added to a volunteer map and recorded a website for, which skews towards the established and the visible. The businesses missing from it are, if anything, likelier to have a worse website than the ones in it, so every figure here is better read as a floor than as an average.

We did not measure Core Web Vitals. Google ranks on field data gathered from real visitors, and no crawler can produce that from a single fetch. The speed figure here is server response time, which is a real number and a smaller one.

We fetched one page per site, the home page. A business with a poor home page and an excellent menu page scores as a poor site. We also measured once, on one day, from one place, so a site that was down for an hour is recorded as down.

The trade categories come from OpenStreetMap tags and are only as good as the tagging. "Shops" and "Offices and agencies" in particular are broad buckets holding businesses that have little to do with each other.

Questions people actually ask

How many business websites in the Gulf are broken?

In this sample of 1,492 businesses across Dubai, Abu Dhabi, Sharjah, Riyadh and Jeddah, 27.8% did not serve a working page: 288 never answered and 127 returned an error, after being retried three times including once as a browser.

What percentage of Gulf businesses publish machine-readable opening hours?

15.9% of the 1,030 working sites we measured. Among the 672 businesses that had published their opening hours to OpenStreetMap, only 18.5% also stated them readably on their own website.

Do businesses in the Gulf block AI crawlers like GPTBot?

Very few. 2.9% block at least one of GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot or Google-Extended, and 10.8% publish no robots.txt at all. Permission is not what keeps these businesses out of AI answers.

Which trade has the worst websites?

Restaurants, on almost every measure: 34.4% broken, a median score of 74, 57.9% with no tappable phone number. Cafés score lower still on the median at 70. Clinics and dentists are the best at 84.

Can I use these figures?

Yes, including commercially. The dataset is on this page. A link back is appreciated and not required.

How can I check my own website against this?

The free check runs the identical code on any URL you give it and reports the same twelve measurements. No sign-up.