A question is coming to tender packs that did not exist three years ago: state whether, and for what purposes, artificial intelligence tools were used in preparing this submission. Nothing in the standard Irish or UK machinery requires it yet - but model wording has been drafted for adoption, and any buyer can insert a clause of their own tomorrow. Bid teams meeting one freeze - is “yes” a mark against us? Is “no” even credible? Here is how to read the clause, how to answer it, and where the practice is heading.
What a well-drafted AI clause actually asks
The disclosure model BidReview has proposed - with model wording drafted for adoption - has one design principle: ask, don’t score. A properly constructed clause does three things. It asks tenderers to indicate whether AI tools were used and, in general terms, for what purposes (document analysis, drafting support, compliance checking). It states explicitly that the answer is for information only and will not be evaluated, so that using or not using AI neither advantages nor disadvantages any tenderer. And it relocates the assurance where it belongs: by submitting, the tenderer confirms that all statements of fact are accurate, that all described experience, personnel and certifications are real and available as described, and that a named officer has reviewed and stands behind the entire submission - whatever the method of production.
Read that way, the clause is not a trap. It is the buyer converting an unmanageable question (which tools did you use?) into a manageable one (who stands behind this content?). The named human is the only assurance that matters.
Why honest disclosure is the safe answer
A false declaration is the real exposure. No procurement rule disqualifies AI-assisted bids - but misrepresentation is a discretionary exclusion ground in Ireland and the UK, and “no AI was used” is increasingly a checkable claim. Drafting residue, template batches and generation artefacts are visible in submitted documents and even in published clarification registers. Declaring “no” and being contradicted by your own document is the one outcome worse than any honest answer.
Scored-disclosure clauses punish honesty - flag them. Some buyers will get this wrong and draft clauses that score the answer or signal that AI use is disfavoured. That design produces routine false declarations and a market taught to hide its methods. If you meet one, it is a legitimate clarification question: ask the authority to confirm whether the response to the AI declaration is evaluated, and how. Their answer is published to all bidders and binds the evaluation.
Disclosure runs both ways. The same logic is coming to the buyer’s side: authorities stating in the tender documents whether and where AI will assist evaluation - compliance checking is different from scoring support, and bidders are entitled to know which. US contractors are currently trying to extract that transparency through litigation; a tender document can give it for free. When you see a buyer-side AI statement in a pack, that is a well-run competition telling you how it works.
How to answer, in practice
Answer factually and generally, matching the clause’s own categories: “AI tools were used for document analysis and drafting support; all content was reviewed and verified by [role], who confirms the accuracy of every statement of fact.” Then make the confirmation true - which is not a form of words but a verification process: claims traced to sources, residue purged, a named person who has actually read the finished bid. The declaration is only as safe as the checking behind it.
The takeaway
Disclose honestly, generally, and without anxiety: the clause done properly is not scoring your tools, it is anchoring accountability for your content. The bid that has something to fear from an AI declaration is not the one that used AI - it is the one nobody read before it went in.
Make the declaration true before you sign it. BidReview audits your finished bid - every claim, every residue risk - so the named person standing behind it can do so with confidence. Run the free scorecard →
