Check my venue →
W11 / Notting Hill · 19 September 2026
We checked 100 Notting Hill restaurants. Here’s what we found.
In September 2026 I checked the information behind 100 W11 restaurants with 50 or more Google reviews: menus, prices, hours, booking routes and signs of activity. Here are the findings.
56%
had a PDF, image or no menu found
40 of 71 assessed websites
51%
had no prices found on their site
41 of 81 assessed websites
40%
had hours flagged as different from Google
34 of 84 assessed websites
11%
used OpenTable as a booking platform
Of 81 assessed websites; preliminary result
Of those I could check. Figures rounded; preliminary results, not fully checked by hand. This research checks online information, not actual AI recommendations.
Your venue / Your result
Want your own result?
Tell me the venue. I’ll send you its score out of 5 and the three things holding it back, free, within two working days. Any London restaurant, bar or café, not just W11.
Open the Check my venue form →
Form not loading? Open it here →
Form not loading? Open it here →
We only use your details to send your result. No list, no newsletter.
The findings / Five questions
01
Can they read you?
Of the 71 venues whose menu format I could assess, 42% had menu text on their site and 1% linked to an external platform. 42% used a PDF, 7% a photo, and 7% had no menu found. Together, PDFs, photos and missing menus accounted for 56%.
A PDF can add an interpretation step. This check does not prove that every AI tool cannot read it. Publishing dishes and prices as web text makes the information easier to access.
Of 81 venues I could check, 25% showed prices as text, 25% showed some prices or kept them in a PDF, and 51% had none found. Percentages are rounded. Missing prices make budget questions harder to answer.
02
Do your details agree?
Of 84 venues I could compare, 33% had website hours matching Google, 40% had a difference flagged, 20% had no website hours found, and 6% stated part of their hours consistently. Different kitchen and venue opening times may explain some flagged differences; a flag is not proof of an error.
For 86% of the 100 venues, the website domain matched the one on Google. The remaining records need checking: another domain can be a booking or ordering page, rather than an error.
03
Do others vouch for you?
75% of the 100 venues had at least one mention from the fixed list of publishers I checked, including Time Out, Hot Dinners, the Standard and guides. 58% had two or more.
That is evidence of coverage. It does not measure whether an AI tool trusts a publisher, or whether every mention is current or favourable.
04
Can they book you?
74% of the 81 websites I could assess had a booking link. This preliminary figure records the presence of a link, not whether the booking journey worked.
A link gives customers a route to book. This audit did not test completed bookings or whether an AI tool could make one. Some venues operate with walk-ins only.
05
Are you clearly still open?
Among the Google reviews returned for the 100 venues, 18% had a review dated within 30 days, 65% within one to three months and 17% had none newer than three months. The median gap was 56 days.
Google’s API returns up to five reviews, not necessarily the latest. An older returned review is not evidence that a restaurant is quiet or closed.
06
Under the bonnet
Of 85 sites probed, 71% carried some of the behind-the-scenes labels on a site. 42% declared a restaurant or related food/local business type, 29% included hours and 14% included a menu reference. These technical checks describe information present in the site code, not how an AI assistant uses it.
These checks can point to content or settings changes. The work needed depends on the website; a rebuild is not assumed.
The scores
A starting point. Not a verdict.
12 venues scored 1, 28 scored 2, 33 scored 3, 23 scored 4, and four scored 5. The average was 2.8. The score describes information foundations under the same rubric for each venue.
12
1
28
2
33
3
23
4
4
5
Score out of 5 · Number of venues shown above each bar
What this means
Start with the basics.
A readable menu, prices, consistent hours and a route to book give customers information they can use.
Some fixes are content or listing updates. They can help people using your website and Google Maps as well as AI tools. No recommendation or booking is guaranteed.
If you run a restaurant and want to know where you stand, the form below gets you your own score, free, in two working days. Individual results are private; only the aggregate is published.
Method and limits
What sits behind the numbers.
Who’s in
The sample contains 100 W11 restaurants with at least 50 Google reviews on 19 September 2026. A later search found 19 more eligible venues, so it covers about 84% of the identifiable pool—not every eligible venue.
What I checked
Menu format, prices, website hours against Google, booking links and platforms, press coverage from a fixed publisher list, the behind-the-scenes labels on a site, content available without running scripts, and the newest review returned by Google’s API.
How I scored
The fixed workbook rubric combines six checks into ten possible points, then groups them into scores from 1 to 5. This is an information-readiness score, not a test of actual AI recommendations or lost bookings.
Where it can be wrong
14 venues had no fetchable website: seven social-only, five unreachable and two not found. Their scores rely on Google. The review sample may omit newer reviews. Instagram posting frequency was checked by hand. The workbook labels 86 calculated scores as final and 14 as provisional. These labels do not mean all results have been checked by hand.
Opening hours versus service times
The hours comparison does not establish whether each listing describes venue opening times or kitchen service times. Treat flagged differences as points to investigate, not confirmed errors.
What I don’t publish
Individual results. If you want yours, ask. Only aggregate findings appear here.
Source: Discovery is moving — W11 AI-Readiness Index v0.4. Fieldwork: 19 September 2026. Preliminary findings; read the method and limitations above.
Your venue / Your result
Want your own result?
Tell me the venue. I’ll send you its score out of 5 and the three things holding it back, free, within two working days. Any London restaurant, bar or café, not just W11.
Open the Check my venue form →
Form not loading? Open it here →
Form not loading? Open it here →
We only use your details to send your result. No list, no newsletter.
Want to know what the audit covers and what it costs? →
Red Hills Lab · London · © 2026
PrivacyPolicy approval pending

W11 / Notting Hill · 19 September 2026

We checked 100 Notting Hill restaurants. Here’s what we found.

In September 2026 I checked the information behind 100 W11 restaurants with 50 or more Google reviews: menus, prices, hours, booking routes and signs of activity. Here are the findings.

56%

had a PDF, image or no menu found

40 of 71 assessed websites

51%

had no prices found on their site

41 of 81 assessed websites

40%

had hours flagged as different from Google

34 of 84 assessed websites

11%

used OpenTable as a booking platform

Of 81 assessed websites; preliminary result

Of those I could check. Figures rounded; preliminary results, not fully checked by hand. This research checks online information, not actual AI recommendations.

Your venue / Your result

Want your own result?

Tell me the venue. I’ll send you its score out of 5 and the three things holding it back, free, within two working days. Any London restaurant, bar or café, not just W11.

Open the form in a new tab →

Form not loading? Open it here →

We only use your details to send your result. No list, no newsletter.

The findings / Five questions

Can they read you?

Of the 71 venues whose menu format I could assess, 42% had menu text on their site and 1% linked to an external platform. 42% used a PDF, 7% a photo, and 7% had no menu found. Together, PDFs, photos and missing menus accounted for 56%.

A PDF can add an interpretation step. This check does not prove that every AI tool cannot read it. Publishing dishes and prices as web text makes the information easier to access.

Of 81 venues I could check, 25% showed prices as text, 25% showed some prices or kept them in a PDF, and 51% had none found. Percentages are rounded. Missing prices make budget questions harder to answer.

Do your details agree?

Of 84 venues I could compare, 33% had website hours matching Google, 40% had a difference flagged, 20% had no website hours found, and 6% stated part of their hours consistently. Different kitchen and venue opening times may explain some flagged differences; a flag is not proof of an error.

For 86% of the 100 venues, the website domain matched the one on Google. The remaining records need checking: another domain can be a booking or ordering page, rather than an error.

Do others vouch for you?

75% of the 100 venues had at least one mention from the fixed list of publishers I checked, including Time Out, Hot Dinners, the Standard and guides. 58% had two or more.

That is evidence of coverage. It does not measure whether an AI tool trusts a publisher, or whether every mention is current or favourable.

Can they book you?

74% of the 81 websites I could assess had a booking link. This preliminary figure records the presence of a link, not whether the booking journey worked.

A link gives customers a route to book. This audit did not test completed bookings or whether an AI tool could make one. Some venues operate with walk-ins only.

Are you clearly still open?

Among the Google reviews returned for the 100 venues, 18% had a review dated within 30 days, 65% within one to three months and 17% had none newer than three months. The median gap was 56 days.

Google’s API returns up to five reviews, not necessarily the latest. An older returned review is not evidence that a restaurant is quiet or closed.

Under the bonnet

Of 85 sites probed, 71% carried some of the behind-the-scenes labels on a site. 42% declared a restaurant or related food/local business type, 29% included hours and 14% included a menu reference. These technical checks describe information present in the site code, not how an AI assistant uses it.

These checks can point to content or settings changes. The work needed depends on the website; a rebuild is not assumed.

The scores

A starting point. Not a verdict.

12 venues scored 1, 28 scored 2, 33 scored 3, 23 scored 4, and four scored 5. The average was 2.8. The score describes information foundations under the same rubric for each venue.

Score out of 5 · Number of venues shown above each bar

What this means

Start with the basics.

A readable menu, prices, consistent hours and a route to book give customers information they can use.

Some fixes are content or listing updates. They can help people using your website and Google Maps as well as AI tools. No recommendation or booking is guaranteed.

If you run a restaurant and want to know where you stand, the form below gets you your own score, free, in two working days. Individual results are private; only the aggregate is published.

Method and limits

What sits behind the numbers.

Who’s in

The sample contains 100 W11 restaurants with at least 50 Google reviews on 19 September 2026. A later search found 19 more eligible venues, so it covers about 84% of the identifiable pool—not every eligible venue.

What I checked

Menu format, prices, website hours against Google, booking links and platforms, press coverage from a fixed publisher list, the behind-the-scenes labels on a site, content available without running scripts, and the newest review returned by Google’s API.

How I scored

The fixed workbook rubric combines six checks into ten possible points, then groups them into scores from 1 to 5. This is an information-readiness score, not a test of actual AI recommendations or lost bookings.

Where it can be wrong

14 venues had no fetchable website: seven social-only, five unreachable and two not found. Their scores rely on Google. The review sample may omit newer reviews. Instagram posting frequency was checked by hand. The workbook labels 86 calculated scores as final and 14 as provisional. These labels do not mean all results have been checked by hand.

Opening hours versus service times

The hours comparison does not establish whether each listing describes venue opening times or kitchen service times. Treat flagged differences as points to investigate, not confirmed errors.

What I don’t publish

Individual results. If you want yours, ask. Only aggregate findings appear here.