How long does it actually take to get an answer? Less than a quarter within the legal limit
The Freedom of Information Act gives public bodies 20 working days to respond. We can now check how often that actually happens: wherever a response states the date the original request or letter arrived, we can pair it with GOV.UK's own publication date and count the working days between them. Across 455 policing & justice releases where both dates could be reliably extracted, the answer is blunter than we expected.
The headline number
Only 22.4% of the requests we could measure were answered within the statutory 20 working days. The median response took 44 working days — more than double the legal limit. A quarter took over 90 working days, the better part of a working year measured in FOI time.
This isn't a cherry-picked worst case. We extracted these numbers the same way throughout: searching response text for an explicit statement of when the original request arrived ("email of [date]", "received on [date]"), pairing it with the release's actual publish date, and discarding anything where the maths didn't make sense (a negative gap, or over a year, which usually means we've matched the wrong two dates). That left 455 genuinely paired requests out of a much larger starting pool - most releases don't restate a date clearly enough for us to extract one automatically, which is itself worth being upfront about below.
It's gotten worse, and not gradually
Splitting by the year the request was made shows a clear, ugly trend, not just year-to-year noise:
| Requested in | Median response | Within 20 working days |
|---|---|---|
| 2012 | 21 working days | 48.8% |
| 2014 | 25 working days | 32.8% |
| 2016 | 57 working days | 11.2% |
| 2018 | 90 working days | 11.5% |
| 2020 | 58 working days | 17.1% |
| 2022 | 101 working days | 6.3% |
| 2023 | 116 working days | 8.9% |
In 2012, requests answered within the legal deadline were close to a coin flip. A decade later, fewer than 1 in 10 were. The middle of that range - roughly 2015 to 2017 - is where the steepest single drop happens, well before the pandemic could plausibly be the whole explanation.
How this compares to the official picture
The Cabinet Office publishes its own annual Freedom of Information statistics, covering the 41 major central government departments and agencies it formally monitors. Their numbers also show decline, but a much gentler one: 87% answered in time in 2020, 88% in 2021, 86% in 2022, 81% in 2023.
Our number for the same period is far lower than that, and the gap is worth explaining honestly rather than glossing over. The Cabinet Office figure covers a curated list of 41 well-resourced bodies with dedicated FOI teams - the Home Office, the Ministry of Defence, and similar. Our dataset instead spans everything GOV.UK's FOI feed has published since 2009, which includes many smaller, less-resourced bodies never part of that monitored list, and our extraction only succeeds on responses that explicitly restate a date - which may itself skew toward slower, more apologetic responses that feel the need to account for the delay. Two genuinely different populations, measured two different ways, are not the same statistic. What both agree on, independently, is the direction: FOI response times in the bodies we can measure have been getting worse for a decade, not better.
It's also consistent with what the regulator itself is now doing about it. In December 2025, the Information Commissioner's Office began requiring councils to submit action plans to reach 90% timeliness and clear overdue-request backlogs by March 2026, after FOI complaints to the ICO rose more than 40% in a single year. That's a regulator responding to the same deterioration our own numbers show, from the enforcement side rather than the data side.
Reading this responsibly
This is a genuinely experimental measurement, not an audited statistic - we extracted dates with pattern-matching against free text, not a structured field either GOV.UK or the responding body provides. Some error rate in the extraction is likely; we mitigated it by discarding implausible pairs and spot-checking a random sample by hand, but we haven't had every one independently verified. Treat the trend - clearly worsening over a decade, corroborated by a differently-measured official statistic moving the same direction - as the more solid finding, and the exact percentages as an informed estimate rather than the last word.