AI-Generated Fake Citations Don't Invalidate Judicial Orders If Correct Law Applied: Andhra Pradesh High Court

Introduction

In a landmark ruling that underscores the cautious integration of artificial intelligence (AI) into judicial processes, the Andhra Pradesh High Court has held that the mere inclusion of non-existent citations generated by AI tools does not automatically vitiate a judicial order, provided the underlying legal principles applied are correct and appropriately tailored to the case facts. Justice Ravi Nath Tilhari, in dismissing Civil Revision Petition No. 2487 of 2025, emphasized the need for "actual intelligence" over "artificial intelligence" while affirming the trial court's decision to retain an advocate commissioner's report as evidence. This decision arises from a property dispute suit where the trial judge inadvertently relied on AI-fabricated case references. The petitioners, defendants in the original suit, challenged the order claiming it was tainted by these fake citations. The ruling not only resolves the immediate dispute but also provides guidance on AI's role in legal drafting, highlighting risks to judicial integrity without outright prohibiting its use.

The case, involving petitioners Gummadi Usha Rani and another against respondents Sure Mallikarjuna Rao and another, revolves around a civil suit for permanent injunction (O.S. No. 773 of 2019) pending before the V Additional Junior Civil Judge, Vijayawada. By integrating insights from recent discussions on AI in Indian and international jurisprudence, this decision reinforces the judiciary's commitment to verifiable legal reasoning amid growing technological adoption.

Case Background

The origins of this legal skirmish trace back to a property dispute in Vijayawada, where the respondents (plaintiffs) filed O.S. No. 773 of 2019 seeking a permanent injunction and related reliefs concerning a plaint schedule property. The petitioners, as defendants 1 and 2, contested the claims, leading to procedural battles over evidence gathering.

In 2023, the plaintiffs moved I.A. No. 178 of 2023 seeking appointment of an advocate commissioner to identify and demarcate the disputed property, particularly in relation to a sale deed executed by the first defendant under Document No. 6404 of 1989. The trial court initially rejected this application on May 2, 2023, prompting the petitioners to file C.R.P. No. 1658 of 2023 before the High Court. On December 12, 2024, the High Court set aside the rejection and directed the trial court to appoint a commissioner, who was to take assistance from a Town/Mandal Surveyor to ascertain whether the plaint schedule property fell within the 1989 document's boundaries.

The advocate commissioner duly inspected the site and submitted a report dated July 7, 2025 (noted as 08.07.2025 in some references). Dissatisfied, the petitioners filed I.A. No. 457 of 2025 under Section 151 of the Code of Civil Procedure (CPC), alleging collusion between the commissioner and plaintiffs, non-compliance with High Court directions (e.g., failure to involve the surveyor), and overall bias. The plaintiffs countered, denying any misconduct and accusing the petitioners of suppressing facts about property divisions and prior sale deeds.

The trial court, in its order dated August 19, 2025, framed three key issues: maintainability of the application, whether the commissioner exceeded the High Court's scope, and substantiation of collusion allegations. Rejecting the petitioners' plea, the court held the report as a valuable evidentiary aid, to be weighed at trial alongside other evidence. It stressed that mere assertions of bias lacked concrete proof and could be tested through cross-examination. Notably, the order cited four non-existent judgments—Subramani v. M.Natarajan ((2013) 14 SCC 95), Chidambaram Pillai v. SAL Ramasamy (1971 (2) SCC 68), Lakshmi Devi v. K.Prabha ((2006) 5 SCC 551), and Gajanan v. Ramdas ((2015) 6 SCC 223)—later revealed to be AI-generated.

Challenging this under Article 227 of the Constitution, the petitioners argued in the revision petition that reliance on fictitious rulings invalidated the entire order. The High Court, on September 26, 2025, sought a report from the trial judge, who admitted using an AI tool for the first time, believing the citations genuine, and pledged future verification from authoritative sources.

This procedural history illustrates the evolving challenges in civil litigation, where local investigations under Order XXVI CPC play a pivotal role in property disputes, and the intrusion of technology into judicial writing.

Arguments Presented

The petitioners, represented by Sri M. Venkata Siva Teja, centered their challenge on the foundational flaw in the trial court's order: its reliance on non-existent citations. They contended that these "fake rulings," generated by AI without verification, misrepresented legal precedents, rendering the order unsustainable. Specifically, they argued that the trial court's acceptance of the commissioner's report—without addressing alleged non-compliance with High Court directions or collusion—was based on erroneous legal propositions drawn from phantom authorities. This, they claimed, violated principles of natural justice and warranted setting aside the order under the supervisory jurisdiction of Article 227.

In contrast, the respondents (though no counsel is noted for them in the judgment) had previously countered in the trial court by denying collusion and highlighting the petitioners' suppression of material facts, such as the first defendant's property division and execution of sale deeds. They maintained that the commissioner's actions aligned with the mandate to identify the property, and any objections to the report should be resolved at trial through evidence, not preemptive striking down.

The High Court itself played a pivotal role by probing deeper into the AI issue. Upon receiving the trial judge's report, the court framed the core question: whether the impugned order deserved reversal solely due to AI-generated citations. The analysis shifted from procedural lapses to substantive legal correctness, with the court emphasizing that evidentiary disputes over commissioner reports must be adjudicated on merits, not ancillary citation errors.

These arguments highlight a tension between technological efficiency and judicial rigor, echoing broader concerns in legal practice about unverified AI outputs misleading courts.

Legal Analysis

Justice Tilhari's judgment meticulously dissects the interplay between AI tools and judicial decision-making, while grounding the ruling in established civil procedure principles. At its core, the decision interprets Order XXVI Rule 10 CPC, which governs commissioner reports as evidentiary aids rather than binding adjudications. The court clarified that such reports form part of the record, subject to scrutiny, cross-examination, and judicial weighing against other evidence—but they do not ipso facto prove title, possession, or misconduct without substantiation.

Drawing on Supreme Court precedents, the judgment invokes Bandhua Mukti Morcha v. Union of India (1984) 3 SCC 161, where the commissioner report was deemed prima facie evidence, open to challenge via affidavits, with courts retaining discretion on its weight. This precedent underscores that reports aid elucidation of disputed matters, such as property boundaries, without delegating judicial functions. Similarly, M.P. Rajya Tilhan Utpadak Sahakari Sangh Maryadit v. Modi Transport Service (2022) 14 SCC 345 elaborates that commissioners perform ministerial, not judicial, acts; their opinions are non-adjudicatory, allowing courts to confirm, vary, or set aside reports post-objections. Paragraph 37 of this case, quoted extensively, reinforces that reliance on such reports merely assists, not binds, the court.

Local precedents further bolster this: In Kanchi Subbamma v. Mannepalli Penchalaiah (1977 SCC OnLine AP 115), the Andhra Pradesh High Court limited commissioner reports to observations of site conditions, lacking substantive evidentiary value beyond corroborating party-led evidence. The Allahabad High Court's Chandrapal v. Roop Ram (1978 SCC OnLine ALL 819) adds that reports must be evaluated alongside other evidence, with commissioner opinions not conclusive.

On the AI front, the judgment innovatively addresses emerging risks, citing international and domestic cases. The English High Court's Venkateshwarlu Bandla v. Solicitors Regulation Authority (MANU/UKAD/0272/2025) condemned fake citations as threats to process integrity, noting AI's role in fabricating authorities without verification. Similarly, Frederick Ayinde v. London Borough of Haringey (MANU/UKAD/0304/2025) held lawyers accountable for diligence, referring solicitors for misleading courts with fictitious quotes. The Bombay High Court's Mr. Deepak v. Heart & Soul Entertainment Ltd. (2026 BHC-AS 828) welcomed AI for research but mandated cross-verification, criticizing unverified outputs like ChatGPT-generated submissions.

Domestically, the Supreme Court's Annaya Kocha Shetty v. Laxmibai Narayan Satose (2025 SCC OnLine SC 758) cautioned against AI-generated statements disorienting cases, advocating technology's efficiency only with human oversight. Justice Tilhari synthesizes these to warn of harms: wasted resources, diverted judicial time, reputational damage, and eroded trust. He advocates "great care, caution and wisdom" in AI use, preferring human intelligence to ensure decisions rest on authentic law.

Crucially, the court distinguishes between substantive errors (e.g., incorrect law application) and formal ones (e.g., citation mistakes). Here, despite fake citations, the trial court's principles—treating the report as weighable evidence, requiring proof of bias at trial—aligned with precedents, causing no prejudice. This nuanced approach differentiates quashing for legal faults versus harmless errors, promoting efficiency without compromising accuracy.

Integrating broader context from recent reports, such as the Andhra Pradesh High Court's own observations on AI in orders and the Delhi High Court's emphasis on scientific evidence over hostile witnesses in unrelated POCSO cases, this ruling signals a judiciary adapting to tech while prioritizing verification. It clarifies that AI hallucinations (fabricated facts) do not per se invalidate orders if reasoning holds, but demands rigorous scrutiny to prevent miscarriages.

Key Observations

The judgment features several pivotal excerpts that encapsulate the court's stance on AI and procedural evidence:

• On the non-vitiating effect of fake citations: "mere mention of the non-existent citations/rulings generated by Artificial Intelligence in the order would not vitiate the order if the law as considered in the order is the correct law of the land and there is no fault in applying the correct law, correctly to the facts of the case."

• Emphasizing human oversight: "the exercise of actual intelligence over artificial intelligence should be preferred and the use of Artificial Intelligence should be done with great care, caution and wisdom. The learned Trial Courts while using the Artificial Intelligence tools in judgments shall remain vigilant cautious and act with judicial application of mind to make just decision ensuring that the judgments/orders are based on correct legal principles."

• On commissioner reports' nature: Quoting M.P. Rajya Tilhan Utpadak Sahakari Sangh Maryadit (supra): "The Commissioner so appointed does not strictly perform a 'judicial act which is binding' but only a 'ministerial act'. ... The Commissioner's report is only an opinion or noting... Such a report does not automatically form part of the court's opinion..."

• Warning on AI risks: "This highlights the risks involved in relying on AI tools without meaningful human oversight to verify the accuracy of the assertions they generate. ... AI systems can produce responses that appear persuasive yet are factually or legally incorrect."

These observations, drawn verbatim, highlight the balance between innovation and integrity.

Court's Decision

The Andhra Pradesh High Court dismissed the Civil Revision Petition, upholding the trial court's order dated August 19, 2025. Justice Tilhari concluded that no illegality or jurisdictional error marred the impugned order, as the applied legal principles comported with established precedents under Order XXVI CPC. The petitioners' opportunity to challenge the commissioner's report at trial via objections and cross-examination was preserved, ensuring no prejudice.

Practically, this means the suit proceeds with the report on record, subject to final adjudication. No costs were imposed, and pending interlocutory applications were closed.

The implications are profound for legal practice. It affirms that technical citation errors from AI do not doom orders if substantive law is sound, encouraging cautious AI adoption for tasks like summarization while mandating verification from official databases. For judges and lawyers, it serves as a cautionary tale: AI aids efficiency but cannot replace discernment, potentially reducing frivolous challenges to orders on peripheral grounds.

In future cases, courts may more readily scrutinize AI-generated content, possibly leading to guidelines on tool usage. This could influence bar council ethics rules, emphasizing diligence in research. Broader effects include bolstering trust in tech-assisted judiciary, as seen in parallel developments like the Delhi High Court's reliance on DNA evidence in POCSO convictions despite witness hostility—both underscore evidence's primacy over procedural hiccups.

Ultimately, this ruling navigates the AI-judiciary frontier, ensuring technology enhances, not undermines, justice delivery. As Indian courts grapple with digital tools, decisions like this will shape a more resilient legal ecosystem, potentially inspiring uniform protocols across high courts.