What impact is AI having on planning systems?

Artificial intelligence (AI) is already embedded in the planning process on both sides of the table. 

Developers are using it to help draft planning statements and collate environmental assessments. Residents, on the other hand, are leveraging it to generate objections. It's also being used as an efficiency tool for local planning authorities processing these cases. 

In the middle of this sit planning surveyors, advising clients on development risks and the quality of documents that end up in front of the decision-makers.

The key question, as in many professions, is not whether AI will affect and reshape planning systems, because it already has. The main issue will be what the use of this technology in these processes means for the obligations, risks and liabilities for all stakeholders involved, and how to manage them. 

Particularly in first drafts, you can see an increase in AI assistance. Access to AI gives the public far more scope to develop and submit these quickly. 

For development surveyors and planning consultants, efficiency is fundamental. Even in larger practices, AI is being used to process technical drawings, identify policy issues and prepare responses for pre-application advice. 

For smaller companies, the technology has an even more obvious appeal, as they can potentially now avoid expensive consultancy fees. 

Lastly, local planning authorities are no different, and it is increasingly common to see the integration of AI in their activities. 

Earlier this year the government announced that English councils would be trialling an AI tool built by Google to help process planning applications. It was framed by Housing and Planning Minister Matthew Pennycook MP as a solution to the major delay in reaching the government's target of building 1.5m homes. 

Interestingly, this was also framed in the Financial Times as a counter to the flood of AI-generated objections taking up officers' resources. Whether this is a legitimate solution to the issue, and what the effect will be of AI being pitted against itself, is not clear yet, but it will undoubtedly have both intended and unintended consequences.  

Naturally, similar tools have followed, and it's easy to make the argument that they are not doing anything unreasonable. Historically, applicants had access to professional consultants while residents did not, creating a clear asymmetry that AI has reduced. 

The consequences for the planning system are worth considering, particularly whether the wider impact of this democratisation of access will be positive or negative. 

It's now common for planning officers to receive large volumes of nearly identical letters containing wide-ranging legal claims with little focus on, or even connection to, the specific planning proposals. 

Managing director at the Strategic Land Group Paul Smith has commented on this new element of the planning process, stating that individual schemes can now bring in thousands of lengthy objections at a time, all made using similar templates.

Objection levels carry significant political weight with elected members of councils, even if the content lacks substantive arguments, meaning that genuine objections and local representatives are being drowned out. 

For development surveyors advising clients, this matters. Facing the challenge of a coordinated campaign of AI-generated objections is now a real consideration that will need to be included in pre-application risk assessments. 

Understanding how a locality is likely to respond to planning processes, and how a planning authority will judge incoming objections, is rapidly becoming a part of the professional advice that clients need. 

By June this year, a tracker made by barrister Matthew Lee had recorded more than 66 confirmed or suspected cases of AI-hallucinated citations, in processes ranging from the High Court through to the First-tier Tax Tribunal. 

Two cases in particular highlight this risk. In Ayinde v London Borough of Haringey [2025] EWHC 1383 (Admin), the High Court noted that fabricated case authorities were produced by AI in proceedings related to housing decisions. 

In Taiwo v Homelets of Bath Ltd [2025] EWHC (KB) 3173, the court found that a fabricated authority falsely created by AI had added significantly to the burden on the court and respondent. 

While neither of these examples is a planning case, the precedent is clear. With the level of AI involvement by all parties at all stages of planning applications, the risk of AI hallucination is extremely high and carries major financial and legal consequences for those involved. 

Large language models generate authoritative-sounding content whether the underlying information is accurate or not. 

As highlighted in the 2024 RIBA AI report, software providers carry no legal liability for inaccurate or hallucinated AI outputs, meaning the practitioner retains total professional exposure.

The same principle applies to RICS members who are preparing or reviewing AI-assisted submissions; the duty to verify is not on the software.

Any use of AI to create or alter material must disclose exactly how it was applied, identify the tool and declare responsibility for the accuracy of the information. 

The H2 Teesside examination is an example of this. H2 Teesside is a nationally significant infrastructure project application for a proposed hydrogen production facility in Teesside. In March 2024 the project submitted a development consent order that was accepted for examination the following month. 

During this examination, the inspector wrote to all parties asking if AI had been used in any of their submissions in line with the guidance. 

Earlier versions of AI guidance raised questions about scope, particularly as routine tools such as grammar checkers could technically be classed within the disclosure requirements.

These discrepancies have been noted by councils. For example, North Norfolk District Council has introduced its own policy regarding AI disclosure, which expressly excludes conventional automated tools such as grammar and spell-check functions. 

These inconsistencies point to a broader feeling that the absence of a national statutory framework creates uncertainty for practitioners working across multiple jurisdictions. 

The Law Commission's discussion paper on AI and the law, published in July last year, is the most detailed attempt currently to identify where the law has fallen short. 

Its central focus for practitioners is what the discussion paper has labelled the liability gap: situations where harm caused by an AI tool cannot easily be attributed to any identifiable person across the development chain. 

In a planning context, those gaps remain unresolved. If an AI tool drafts a viability appraisal containing an error that a planning officer relies on when making a decision, existing law does not clearly point to who is at fault. 

Part of the issue is the speed at which AI is evolving; policy updates are part of a far slower machine, and attempts to update or evolve with the technology are proving a major challenge. 

The Law Commission has launched a new project specifically for automated decision-making by public bodies, recognising that judicial review is not well suited to scrutinising automated decisions. 

This may develop and produce a statutory framework, but that remains to be seen. 

Firstly, verification is an obligation and not an option. AI-assisted drafting has a completely legitimate use in planning submissions, but it's crucial that any information it provides or produces is independently checked by a human. 

This includes policy citations, case references and technical figures in any document bearing a surveyor's name, and these should be checked by a qualified professional before it goes further. 

Secondly, disclosure is becoming standard practice, and it is better to go above and beyond the current requirements. The Planning Inspectorate's guidance creates a clear expectation of transparency at all stages, and non-compliance carries significant risk. Keeping a detailed record of the tools you use is a must. 

Development risk assessments should now treat AI-generated objections as a credible risk, especially for contested or high-profile schemes. Surveyors advising before submission should factor in the chance of a high-volume campaign when assessing planning risk and timelines. 

Early community engagement remains the best preparation. It gives residents a clear way to raise concerns and reduces the risk of AI-generated campaigns dominating the consultation. 

A high number of objections can influence committee members, even where the points are weak or repetitive. Clients should be warned about this risk, and committee presentations should focus on the substance of the objections, not just the total number received.

The Law Commission may produce a substantial legislative framework that resolves the current liability gap. However, until it does, chartered surveyors stand between the technology and the decision-maker. 

This is not an easy job, and it is made harder without clear rules. This must be a familiar feeling, as managing uncertainty on behalf of clients is exactly what the profession does. 

The time to get ahead is now. Don't just rely on legislation that may or may not appear; engage in proactive good practice now, as it is inevitable that a significant claim will arise from an AI-assisted planning submission that no one properly verified.