Last updated: 1 November 2026, 09:16 GMT. This report examines Google DeepMind’s 31 October announcement and the government’s 30 October release.
The PlanAssist AI planning trial is expanding to 15 councils in England for 12 months, after Google DeepMind reported that early testing reduced the time officers spent on householder assessment tasks by up to 58%. The next phase uses live applications, with human review retained before decisions.
PlanAssist AI planning: what is actually new
Google DeepMind’s announcement says 12 additional authorities will join the initial pilots in Barnet, Camden and Dorset, in three cohorts. The Gemini-based tool was developed with the UK government, Google Cloud and Faculty.
Its described workflow brings together application documents, drawings, policies and site constraints, checks for missing information and issues, and drafts cited reports for officers to edit. DeepMind says more than 70% of generated reports in early testing were approved without edits. That is a reported review outcome; it is not an independently established error rate.
The housing ministry’s release corroborates the expanded trial and says it intends to make the tool available to all English councils by the end of 2027. Availability should not be read as a confirmed adoption date for an individual authority.
Task time and a planning decision are different measurements
The important question is what the percentage measures. Time spent by an officer on selected assessment work is one part of an application’s journey. A faster draft is useful, but the headline number alone cannot demonstrate the elapsed time between submission and a decision.
The government’s release links the technology to an ambition to halve routine householder processing from eight weeks to four. That is a much broader service outcome than a measured reduction in assessment work. Readers should ask for results from live cases before assuming those two measures move together.
There are several useful tests for the next phase. Does a time saving survive the checking and correction process? Are applications of comparable complexity being compared? Do reports cite the policy that actually applies to the site? Can an officer trace an assertion back to the underlying document? These are editorial questions for evaluation, not claims that the pilot has failed them.
Why it matters for Brighton and Hove
For someone extending a home or commenting on a neighbour’s proposal, the value would be a clear, timely, properly reasoned assessment. Software is not the outcome residents need; a defensible decision is.
The announcements reviewed here do not establish that Brighton & Hove is participating. We are not reporting a local launch or advising applicants to submit through a new AI service.
This trial also needs to be distinguished from the new officer-delegation rules covered in our separate report. Who has authority to decide an application and which software helps prepare its assessment are different questions. One should not be used as evidence that the other has been adopted locally.
What residents should do
Continue to use your council’s published application and consultation routes. Keep the application reference, relevant plans and response deadline together. An announcement about a trial does not change the deadline shown for a live local application or remove the need to provide supporting evidence.
If a council announces adoption, look for its explanation of the tool’s role, human checking, document handling and route for correcting an error. Those details will tell residents more about the service than a general productivity percentage.
Brighton Locals’ view
Editorial judgement: Testing an assistant on live work before wider release is a sensible next step. The promise is credible enough to examine, but the standard should be faster decisions with reliable reasoning, not simply more drafts produced. Published comparisons of time, quality and correction work would make the case stronger.
What to watch next
Watch for the participating authorities, evaluation results from the 12-month phase and evidence supporting any wider rollout. An independent assessment of completed cases would materially improve confidence in the benefits claimed.
Bottom line: The trial is expanding; the proof residents need is a better planning service.
Featured image: archive Googleplex headquarters building and sign, 27 July 2016. The Pancake of Heaven! / Wikimedia Commons, CC BY-SA 4.0. Resized for display under the same licence. It depicts Google’s premises, not PlanAssist, a planning application or a Brighton pilot.
Leave a Reply