Can I Use AI to Write a UK Tender Response? PPN 017 Rules, Plain English

Yes, you can use AI to draft any part of a UK public-sector tender response. The rule is disclosure, not abstinence. Procurement Policy Note 017 (live 24 February 2025, replaced PPN 02/24) explicitly assumes bidders will use AI tools and tells contracting authorities to ask three Annex B disclosure questions about it.

Disclosure answers are typically marked "For Information Only" and not scored. What gets you disqualified is undisclosed AI use, or pasting the buyer's confidential annex into a public LLM as training data. Below: the verbatim Annex B questions, what counts as compliant AI use, and the practical 6-step checklist.

• Yes, you can use AI on UK public-sector tender responses. PPN 017 (live 24 February 2025) explicitly assumes you will and tells buyers what to ask about it.
• PPN 017 replaced PPN 02/24 (which ran from 25 March 2024) when the Procurement Act 2023 went live.
• Disclosures are "For Information Only". They are not scored as part of the formal evaluation in most cases.
• There are three Annex B questions buyers are encouraged to ask. Get the verbatim wording below and pre-write your answer.
• What gets you disqualified is undisclosed AI use, not AI use itself. Plus uploading confidential government info to a public LLM as training data.
• Practical fix: keep an audit trail of every AI-drafted section, run a human review pass on every word, and answer the disclosure questions honestly.

PPN 017 does not tell you to stop using AI. It tells you to be transparent about it. The bidders panicking about disclosure are losing to bidders who pre-wrote their disclosure answers six months ago.

What's in this guide¶

• The short answer (yes, with disclosure)
• PPN 02/24 vs PPN 017: which one applies to your bid
• The three disclosure questions, verbatim from Annex B
• What you must not do (the actual disqualifier)
• A practical AI-on-tenders checklist for cleaning SMEs
• Where the rule fits with the rest of the Procurement Act 2023

The short answer¶

You can use AI to draft any part of a UK public-sector tender response. The rule is disclosure, not abstinence. Procurement Policy Note 017 (the successor to PPN 02/24) is the live guidance and it actively assumes bidders will use AI. Annex B of the PPN tells contracting authorities to add three example questions to their Invitation to Tender so they can collect disclosure data.

Critically, the disclosure questions are typically marked "For Information Only". They are not formally scored against the award criteria in most procurements. The buyer wants to know what tools you used so they can do due diligence on accuracy and capability, not so they can deduct points.

PPN 02/24 vs PPN 017: which one applies¶

If you are bidding on anything with a tender notice published from 24 February 2025 onwards, PPN 017 is the rule that applies. The Annex B disclosure questions are nearly identical between the two PPNs. The substantive change is the statutory framework underneath, not the disclosure mechanics.

The three disclosure questions, verbatim¶

Annex B of PPN 017 provides three example questions that contracting authorities are encouraged to include in the Invitation to Tender. Pre-write your answers to all three. Reuse them on every bid. The questions are quoted verbatim below from the official guidance.

Question 1: AI use in tender preparation

Have you used AI or machine learning tools, including large language models, to assist in any part of your tender submission? This may include using these tools to support the drafting of responses to Award questions.

PPN 017, Annex B, question 1

If yes, the buyer will ask a follow-up: "Where AI tools have been used to support the generation of Tender responses, please confirm that they have been checked and verified for accuracy: Yes/No". Answer Yes only if you can prove it. A human checked every sentence, you can name the reviewer, you can show timestamps.

Question 2: where exactly AI was used

Please detail any instances where AI or machine learning tools, including large language models have been used to generate written content, or support your tender submission.

PPN 017, Annex B, question 2 (PPN 02/24 wording is near-identical)

List the sections. "AI used to draft initial responses for sections 4.1 (method statement), 4.3 (social value commitments), and 4.5 (mobilisation plan). All output reviewed and edited by [name], [role], on [date]. Final wording is the bidder's own." That answer is enough. The buyer is not looking for a research paper, they are looking for an audit trail.

Question 3: AI in service delivery

Are AI or machine learning technologies used as part of the products/services you intend to provide to [Insert Contracting Authority Name]?

PPN 017, Annex B, question 3

This is a different question. It asks whether the actual cleaning service you are delivering will use AI (rota optimisation software, IoT cleanliness sensors, AI-driven scheduling). For most soft FM SMEs the honest answer is No. If Yes, the follow-up asks: "Please describe how AI technologies are integrated into your service offerings. Please provide details: ………………".

What you must not do¶

PPN 017 is permissive on disclosure but firm on two things.

1. Do not lie on the disclosure. If you used AI and you tick No, that is a misrepresentation. Misrepresentation in a procurement response is a discretionary exclusion ground under the Procurement Act 2023. The bid can be rejected and you can be excluded from future bids.
2. Do not paste confidential government information into a public large language model as training data. PPN 017 explicitly tells bidders to make sure they do not use confidential government information as training data for LLMs or AI systems when drafting tender responses. This is the single hardest line in the guidance.
3. Do not submit AI output without a human review. The follow-up to question 1 explicitly asks you to confirm AI output was checked and verified for accuracy. If you cannot say Yes truthfully, you should not be submitting it.

An AI-on-tenders checklist for cleaning SMEs¶

Six steps. Use them on every bid where you bring AI into the workflow.

1. Pre-write your three disclosure answers. Save them as a master document. They change once a year, not every bid.
2. Pick a reviewer. One named person who reads every AI-drafted section before it goes anywhere near the submission. Their name goes in the answer to Question 2.
3. Keep the audit trail. Save the AI prompt, the AI output, and the final reviewed version for each section. A simple shared folder per bid is enough.
4. Do not paste the buyer's confidential annex into a public LLM. If the tender pack is marked confidential, draft the response in a private, contracted AI tool (one with no training-data retention) or write that section by hand.
5. Use AI for drafts and structure, not for facts. AI can give you a great first draft. It cannot tell you the right insurance level or the right BS EN standard. Verify every factual claim against the spec, the marking matrix, and your own records.
6. Read the marking matrix before, during, and after the AI draft. The matrix is what scores. AI does not see it unless you paste it in.

How a purpose-built tender AI tool differs from ChatGPT¶

Generic AI tools were built to write anything. They were not built to keep an audit trail per bid section, hold a compliance vault that contains your real insurance figures and accreditations, or pre-fill the three PPN 017 Annex B disclosure questions. They will happily invent a Companies House number to fill a gap.

CleanTender takes a different shape. Every draft section is grounded in your stored Compliance Vault: the figures, certificates, and case studies you have already verified. If a bid asks for a number we do not have on file, the draft flags it for you to add rather than guessing. Every section is timestamped and saved so you have an audit trail ready for the PPN 017 follow-up question. The three disclosure answers are pre-filled with the named human reviewer and the model used, so the work of staying compliant is done before you open the next tender pack.

It is a different category of tool. ChatGPT is an open-ended writing assistant. A purpose-built tender tool is closer to a contract manager who never forgets the marking matrix and never invents an insurance figure.

AI use that is fine vs AI use that is not¶

How this fits with the Procurement Act 2023¶

PPN 017 sits inside the wider Procurement Act 2023 framework, which went live for new procurements on 24 February 2025. The Act introduced new exclusion and debarment rules for misrepresentation, professional misconduct, and breach of contract. Disclosing AI honestly is a low-risk activity. Lying about it triggers the misrepresentation exclusion grounds.

If the wider Procurement Act 2023 framework is new to you, the plain-English overview is in the Procurement Act 2023 explainer for FM suppliers. For the disclosure machinery in context of the rest of the bid, the Standard Selection Questionnaire walkthrough covers where each question lives in a typical UK public-sector tender pack.

What to do this week¶

1. Pre-write your three PPN 017 Annex B disclosure answers. Keep the document with your master Standard Selection Questionnaire.
2. Pick a named reviewer for AI-drafted sections. Add their name to your master disclosure answer.
3. Set up a shared folder for AI audit trails. One sub-folder per bid. Save prompt + output + reviewed version per section.
4. Audit your current tools. Are any of them using buyer-confidential information as training data? If unsure, switch to a contracted tool or stop using them on confidential annexes.
5. Read the next tender pack you open with the marking matrix in mind, not the page count. AI does not see the matrix unless you paste it in.