Home Office used AI to put together new immigration and settlement consultation
AI tools were used to keep track of and sort the more than 200,000 public consultations to the Home Office’s new immigration and settlement rules.
The UK Government ran a public consultation on proposed changes to the current settlement model from late November 2025 to early February 2026.
The Earned Settlement consultation received 204,667 online responses. Almost all (202,241) were from individuals, with 2,426 from organisations. 904 submissions were received by email.
IFF Research, which handled the survey, confirmed that it used AI tools to help sort and organise the vast amount of data and responses.
“IFF’s in-house coding team developed an initial codeframe which was quality assured by our coding manager and a member of the research team. Separately, two members of the research team also developed an initial codeframe,” the firm said in an attached methodology statement.
“As a check, we used: 1) an in-house AI tool to compare the human codeframes, as well as 2) AI-assisted codeframe development in Ascribe1, for individuals.”
The initial codeframes were run using AI Ascribe, with sequential checking, correcting and quality assurance by our coding and research teams, it said.
“We then ran the corrected codeframe on the same sample, and the coding was again reviewed and corrected by the research team, with Director checks, to improve accuracy, reliability, and validity of the coding. Following further corrections, the codeframe was then run on the full sample of responses, with a further cycle of review and recoding.”
The group noted that due to the high volume of open text responses, it was not possible to manually review every response in full.
“Coding was therefore supported by AI tools, alongside researcher quality assurance checks. Findings from open text responses are reported descriptively in terms of key themes and commonly expressed views,” it said.
AI also used to read email responses
The group noted that the 904 email submissions to the consultation did not follow the same structure as the consultation survey so were analysed separately. This also involved the use of AI tools.
Home Office policy officials analysed the submissions that came with attachments, which were often detailed reports.
“The remaining responses (around 600, or roughly 0.3% of the total consultation responses) were analysed using an in-house tool developed by FF Research’s lead, which identified the top themes,” IFF Research said.
“This analysis and coding were then manually reviewed and Q ’d by two members of the IFF research team. The key themes covered were similar to those in the online responses.”
Two members of the IFF research team then reviewed and checked that analysis by hand.
Despite this, IFF said it reports its findings on the written answers as broad themes rather than precise proportions of respondents.
