AI in planning casework: does the updated guidance go far enough?

The Planning Inspectorate (PINS) updated its guidance on the "Use of artificial intelligence in casework evidence" on 23 September 2026. In our blog below, Doug Haycock considers how the guidance has evolved, the details of the recent update, the implications for the NSIP regime and a comparison with the published position from Planning and Environment Decision Wales (PEDW).

How the guidance has evolved

The original September 2024 guidance asked parties to disclose the AI tools used, the sources the AI relied on, what it created or altered and the date of use. It also required AI-generated content to be clearly labelled in the body of the document.  In February 2026, the guidance was updated to introduce a "golden rule" of responsible use, making each party responsible for the accuracy of AI-generated material. In-body labelling was replaced by a short statement from professional and interested parties. 

What the September 2026 update changes

The guidance now requires the statement to confirm that AI was used, explain what it was used for and how (identifying the relevant paragraphs, sections or images), and confirm what checks were made and that the party takes responsibility for accuracy. Unlike the February version, it no longer needs to name the AI tool. 

The update also adds guidance for particular groups:

• Professional parties: where documents draw on several experts, each expert who used AI should give their own declaration, as a "catch all" declaration is unlikely to be enough;
• Local planning authorities and applicants: where AI is used to summarise representations or consultation responses, including in National Infrastructure examinations, the original documents must still be provided;
• Interested parties: submissions should be as focused and relevant as possible, as overuse of AI can stop the Inspector understanding the main points. This last addition is helpful. It does not, however, take the control of AI much further than the existing position, under which improper use of AI could be treated as unreasonable behaviour and leave that party open to an award of costs. 

PINS' position on NSIPs

PINS’ advice on Nationally Significant Infrastructure Projects (NSIPs) states that where AI has been used to prepare or change a representation, the representation must say where and how. If it does not, the Examining Authority can reject the submission. 

The market now contains several AI specific tools for planning objections. We are also seeing an increase in the use of AI for submissions by interested parties which are invariably extremely lengthy, do not carry a declaration and contain errors of fact. Hopefully, the guidance will help in stopping this.  There is also an increase in the volume of representations being made.  In response, we are seeing an increase in the use of AI to process these submissions to summarise and identify key issues.  Where it used to be the case that one letter would be signed by many AI has made it easier for the same, or very similar letter, to be submitted many times.  It has always been the case that 100 poor arguments will be unlikely to be given much weight, whereas a single good point can have a material effect in the decision-making process. While AI is a useful tool and can enable meaningful participation by those who would otherwise not engage there is a very real question about whether it is adding to the efficiency and effectiveness of proceedings as the approach seems to be moving towards make all the points possible without consideration as to their relevance to the project being determined. 

PEDW’s position

In comparison, PEDW issued guidance on Using AI in submissions on 9 June 2026. With reference to using AI for commenting on a case, the guidance recommends that lengthy generic AI content should be avoided as it is less effective at making compelling arguments and can be unhelpful to the decision maker. As written submissions can be scrutinised at oral hearings and inquiries, it also highlights that submissions may need to be explained and can be questioned, including under cross-examination. It sets out five overarching principles for the use of AI to generate evidence: transparency, accuracy and content integrity, privacy and data protection, copyright and review and oversight. On the latter, it recommends that a person always remains responsible for submissions and that AI outputs are contextually accurate.

Next steps

The guidance from PINS and PEDW is non-statutory and non-binding so it remains to be seen whether it will alter the current practices but everyone involved in the process should consider how they can apply it and take proactive steps to do so.

If you have any queries about the use of AI in relation to the planning regime, please do not hesitate to contact Cathryn Tracey, Jen Ashwell or Doug Haycock.