The claim: Almost all refused asylum seekers are eventually recognised as refugees.

Misuse

appeal success rate generalised to all refusals

Direction

pro-migration

What the data says

In the year ending June 2026, 38% of initial decisions resulted in a grant of protection or other leave. That rate is published on the main-applicant basis.

On the people basis, main applicants plus dependants, 115,872 people received an initial decision, of whom 73,478 were refused. The rate and the count are not two views of one sum: dividing one by the other mixes the bases, which is the error this site warns about most often.

Around 39% of asylum and protection appeals determined at a hearing or on the papers at the First-tier Tribunal in the quarter April to June 2026 were allowed. That is a share of the appeals determined, and only about half of the asylum and protection appeals disposed of in that quarter were determined at all: 37% were withdrawn, 3% struck out and 5% invalid or out of time.

Separately, for claims made between 2007 and 2022, the Home Office's cohort analysis puts the grant rate at the latest outcome 8 to 29 percentage points above the initial rate for the same cohort, a gap the publisher attributes to appeals and reconsiderations, so it is not appeals alone. Claims reinstated after a withdrawal or an administrative outcome may also feed the later figure; that reading is this site's inference, not a link the publisher draws.

So the direction of the claim is right, and the size of the effect is not trivial. A refusal at initial decision is not the end of the matter, and a substantial minority of claims end differently from how they started.

Taken on its own terms, though, the cohort measure does not say "almost all". The Home Office's grant rate at the latest outcome ran from 50% for claims made in 2010 to 75% for 2019, and reached 80% for 2020, the highest of those years. So between a fifth and a half of those claims stood at a refusal rather than a grant.

That rate is not a share of the cohort, and on a page about denominators the difference has to be said out loud. It divides grants by grants plus refusals. It leaves out claims that were withdrawn, given an administrative outcome, or had no outcome recorded yet, and those together run from about a tenth to about a fifth of a year's claims. So measured against every claim made in a year, rather than against the ones settled either way, the share that has not ended in a grant is higher than the figures above, not lower. This page uses the publisher's own measure, which is the more cautious of the two.

What the measure cannot do is arithmetic on the current 38% rate: the Home Office says the two relate to different cohorts of initial decisions.

Why the claim is misleading

It takes the appeal success rate and applies it to the whole refused population, which involves three errors compounding.

First, not every refusal is appealed. The appeal success rate describes the outcomes of appeals that were lodged and then determined. It says nothing about refused claimants who did not appeal, and nobody can assume they have the same profile as those who did.

Second, an appeal allowed is not itself a grant of status. Appeals can succeed on procedural or human rights grounds, and the tribunal does not issue the leave. The Home Office may first seek permission to appeal to the Upper Tribunal; where it does not, its own caseworking team implements whatever form and duration of leave follows, and there is no general target for how long that takes.

Third, protection or other leave is broader than being recognised as a refugee. That category counts refugee permission alongside humanitarian protection, discretionary leave and other grants. A rise in grants of protection or other leave is therefore not the same as a rise in refugee recognition, which is what this claim asserts.

There is also a period problem worth naming. The 38% initial grant rate covers a full year to June 2026. The 39% appeal success rate covers a single quarter. Setting one directly against the other, as though they described the same population over the same period, is not a valid comparison, and it is the comparison the claim implicitly makes.

The mirror claim fails too

"A refusal means the original claim was obviously false" is wrong for the same underlying reason, and it now has a page of its own. A substantial share of refused claims later end differently, which is difficult to reconcile with the idea that a refusal settles the question. Both claims take a real and partial effect and inflate it into a total one.

What follows from the delay

The appeals backlog stood at 90,341 outstanding cases at 30 June 2026, a record high, and the mean age of an asylum appeal at conclusion was 72 weeks. That means a large number of people whose final outcome is not yet known, and it means any figure for "eventual" recognition is provisional, because a large share of the relevant cases has not concluded.

Better questions to ask

• Is this the grant rate at initial decision, or at the latest outcome after appeals and reconsiderations?
• What share of refusals were actually appealed?
• Over what period, and does the appeal figure cover the same period as the decision figure?

Source notes

Initial decisions and grant rate: Home Office, immigration system statistics, year ending June 2026, published 27 August 2026. The published grant rate is calculated on main applicants; the refusal count above is on the people basis, including dependants. The two bases are not interchangeable, which is why the rate and the count are cited separately here rather than divided into one another.

Appeals: HM Courts and Tribunals Service / Ministry of Justice tribunals statistics, April to June 2026, table FIA_3. A different publisher on a different cycle from the Home Office figures, and a single quarter rather than a year. The latest quarter is provisional and liable to revision; the quarter before it is marked revised in the same table.

This page mixes three reference periods, and they cannot be set against one another: a year to June 2026 for the initial decisions, a quarter to June 2026 for the appeals, and cohorts of claims made between 2007 and 2022 for the later grant rates.

The 8 to 29 percentage point gap is a cohort measurement covering claims made between 2007 and 2022, not a current one, and it must not be applied to the latest year to produce a final rate. The Home Office says the initial rate in its cohort table "will not match the grant rate in the 'initial decisions' dataset as they relate to a different cohort of initial decisions", and that more recent cohorts are not comparable because more of their cases are still progressing. Its own terms for the later figure are "latest outcome" and "latest decision"; it does not call it final.

And those outcomes are dated. The cohort table's own note puts them as at January 2026, so anything decided after that, on appeal or on reconsideration, is not yet in the later figure. The bulletin dates the same outcomes to April 2026 instead; January is followed here, being the date the table publishes for itself and the earlier of the two, so this page claims no more currency than the publisher's own table does.

On selection. This site checks propositions that circulate rather than quotations attributed to particular people, as the style guide sets out. For this claim we have not located a published example in these exact words. The versions we have found are weaker, putting it as "many" or "most" rather than "almost all". Correcting the strongest form of a claim when only weaker ones circulate risks answering something nobody says, and this page is published with that objection on the record rather than settled.

All common claims