London Travel Geek
How we rank
Our guides are not a list of places we liked. Every venue is judged on how many independent sources name it — awards, critics, specialists and the people who actually went and filmed it — counted, not asserted. This page explains how that count is built, and where it fails.
Why counting rather than reviewing
A single enthusiastic review is one person on one night. It might be the chef's best service of the year or their worst. What it cannot tell you is whether a kitchen is reliably good, and that is the only thing worth knowing before you travel across a city to eat somewhere.
So we invert it. Instead of adding one more opinion, we record who has already published one, and count the places that separate judges keep arriving at independently. Ten outlets converging on the same address is evidence in a way that any one of them alone is not.
The sources behind this currently number 1,475 source records across 106 topics, carrying 16,862 citations drawn from 562 registered publications, channels and award bodies. Of the venues named in it, 1,874 are named by two or more independent sources and 617 carry a dated award.
The source tiers
Not every mention is worth the same. Sources are registered in advance and tiered by what kind of judgement they represent, so an unknown source is visible rather than silently weighted:
- A — a judged, dated ranking. A named panel or inspectorate published a ranked or starred list with a year attached: 50 Top Pizza, the National Pizza Awards, the Michelin guide. The strongest single signal, because it is falsifiable and re-run annually.
- B — an editorial publication. A masthead with editors paying writers to visit. Broad and useful, but over-represented, and its members read each other.
- C — an independent blog or specialist. One or two people writing from their own visits, or a single-subject specialist. Often earlier and more specific than the publications, and much closer to what a local would actually say.
- D — video. YouTube and TikTok creators who went, filmed and rated. The tier is the medium, not the audience: a channel with two million views and one with twenty thousand are both D.
- E — community. Reddit threads and forum posts where named people argue about where to eat. Counted per thread, never per comment, so one busy thread cannot outweigh a publication.
- F — the venue's own website. Used to check an address, a price or which branches are open. Never counted toward consensus — a restaurant recommending itself is not evidence.
Awards weigh more than press; press weighs more than creators. A venue needs support from two or more sources across two or more tiers, or one tier-A award, before we treat it as properly established. Five publications agreeing is weaker than a publication, a blogger and a video agreeing, because the publications read each other and the three strangers do not.
The counting rules
- Support is counted in distinct domains, never in URLs.Two lists from one publication is one opinion. Time Out's ranked guide, its video channel and its news desk are all Time Out, and count once.
- Syndication credits the original. Where one outlet reports another's ranking — "X has been crowned by the Telegraph" — the source is the Telegraph. Counting both would let a single judgement pose as two.
- Aggregators are excluded entirely. TripAdvisor, Yelp, OpenTable, Wanderlog and the rest rank by user volume, repeat each other and dominate search results. One crowd-sourced ranking can otherwise masquerade as many independent opinions.
- Google ratings are tracked separately and never form part of a placement. Review volume measures footfall, not cooking.
- Age is recorded. A mention older than three years still counts but is flagged; a venue with nothing inside two years is treated as a closure candidate and re-checked.
- Closed venues are dropped even while live guides still carry them — published lists are slow to update and some never do, so the longer a place has been shut, the better supported it can look.
Why there is no score
We could turn all of this into a single number out of ten. We deliberately do not, because the weighting behind it — how much more an award is worth than a magazine, a magazine than a video — is a judgement call, and a decimal point would imply a precision the method does not have. A score invites the question "how was that calculated?" and the honest answer is "by a person deciding".
So we publish the evidence rather than the verdict: how many independent sources named a venue, and where it placed in the lists that actually published an order. You can disagree with our reading and still use the count.
For the same reason, a rank is only ever shown against a source that published one. Plenty of respected guides are deliberately unordered — The Infatuation's best-of guides and Country & Town House both are, and one widely shared social list says "in no particular order" in its own caption. Attaching a position to any of those would be inventing it.
What this method cannot tell you
Counting sources measures agreement, not quality, and the two come apart in predictable ways:
- It rewards being written about. A superb neighbourhood restaurant that no critic has reached will score badly against a mediocre one with a publicist. Central and east London are covered far more heavily than the outer boroughs.
- Some categories have no judged award at all. There is no inspectorate for the full English breakfast or the Sunday roast. Those rankings rest on editorial consensus alone, which is a weaker foundation, and the guide says so where it applies.
- Primary sources are sometimes unreadable. The National Pizza Awards publish their winners as images and their previous-winners page stops at 2024, so the 2025 results are only available through Time Out's reporting. We attribute those to the awards — they are the originating judgement — but the link cannot be checked against a primary page, and we say so in the guide rather than hiding it.
- One venue can be recorded under several names. Sources spell things differently and residencies get filed under their host pub. We merge these deliberately and keep a written note of every merge, but a merge is a judgement and occasionally it will be the wrong one.
- It is a snapshot. Restaurants close, chefs leave and residencies end faster than guides update. Every ranked page carries the date its evidence was last rebuilt.
Independence
No placement on this site is paid for, and no venue can buy its way into a guide or out of one. Where a page carries affiliate or partner links they are marked as such, and they play no part in what is ranked or how.
Corrections
If a ranking here looks wrong, the useful thing to tell us is which source we have missed, miscounted or double-counted — that is a claim we can check and fix, and the count will change with it.