GAO Rebukes Protester Over Hallucinated Citations, Reminds Firms of Risks of AI/LLM-Assisted Legal Practice

A recent decision from the Government Accountability Office (“GAO”) highlights a growing – and increasingly risky – trend in bid protest practice: the continued perfunctory use of artificial intelligence (“AI”) or large language model (“LLM”) tools to assist with protest filings. Protesters should heed this decision because it, once again, reminds industry that GAO reserves the right to issue sanctions where a protester's actions undermine the integrity of its bid protest process. 

In Bramstedt Surgical Inc., B-424064, Jan. 28, 2026, 2026 CPD ¶ __, GAO dismissed an incumbent contractor’s bid protest challenging a procurement for surgical instrument maintenance and repair services. Among other arguments, the protester alleged that the agency failed to directly notify it of the solicitation, depriving the firm of an opportunity to compete for work it was previously awarded. The agency rescinded that award after determining the "instrument list" was inaccurate.

GAO rejected the protester's arguments on familiar grounds, concluding that the protest "fail[ed] to state a legally sufficient basis of protest" and that it raised matters of contract administration over which GAO does not have jurisdiction. Indeed, GAO noted that the agency satisfied the applicable FAR requirements by posting the solicitation on SAM.gov, which put the firm on constructive notice of the solicitation. Standing alone, GAO's dismissal would not have been remarkable.

What makes the decision noteworthy, however, is GAO’s discussion of the legal citations in the protester’s filings, which GAO observed bore the hallmarks of unverified AI/LLM-assisted drafting. In addressing those deficiencies, GAO provided a robust discussion of its expectations regarding the use of AI/LLM tools in bid protest practice.

While GAO ultimately reiterated that there is nothing inherently improper about the use of AI/LLM tools in legal practice, the decision underscores that inaccurate citations may expose protesters to dismissal or sanctions.

GAO Flags Hallucinated Citations

After dismissing the protest on traditional grounds – failure to state a legally sufficient basis and lack of jurisdiction – GAO addressed deficiencies in the protester’s legal citations, observing that several bore the hallmarks of AI/LLM-assisted or -generated output that had not been independently verified.

For instance, GAO explained that one of the protester's citations did not correspond to any bid protest decision with the "B-number, party name, and date combination." GAO further noted that it had issued decisions matching either the B-number or the party name, but that none articulated the legal principle for which the citation was offered and therefore were not relevant to the protest.

In another example, GAO found that the proffered citation was only partially accurate: the party name and B-number matched two different GAO protest decisions issued on different dates. Neither decision, however, stood for the legal proposition on which the protester relied nor were they relevant to any other issues in the bid protest.

GAO cautioned that the use of AI/LLM tools to draft or assist in drafting filings can result in the citation of non-existent cases, and that reliance on such programs without verification for accuracy wastes the time of all involved, including GAO. And while GAO emphasized that there is nothing wrong with properly using AI/LLM to "practice law or litigate cases," it noted that these nascent tools often experience glitches and hallucinations. As a result, “close and careful attorney supervision, fact-checking, and citation-checking are absolute necessities when utilizing AI or any of its subsets.”

GAO concluded by reiterating its inherent authority to dismiss protests and impose sanctions where the "protester's actions undermine the integrity and effectiveness of [GAO's] process.” While GAO declined to impose sanctions – because the protest was dismissed for other deficiencies – GAO made it clear that this authority remains available where a protester submits non-existent case law.

Takeaway

This bid protest decision offers a cautionary reminder that even as advanced technologies like AI/LLM tools reshape the way attorneys and other advocates work, the fundamentals of legal advocacy remain unchanged. Accurate citations, careful analysis, and rigorous verification of legal authority are essential, particularly when AI/LLM tools are used to draft filings or synthesize legal cases. Contractors should therefore take note: failure to verify legal citations – including cited legal principles and quoted language – may not only weaken your bid protest but also, in certain circumstances, result in dismissal or sanctions.

. . .

#govcon