Courts Highlight Limitations with Text and Chat Message Searches
Editor’s Note: A conversation can reveal exactly what a legal team needs to know without containing a single word on its search list. In this article for HaystackID®, Phil Favro examined why text and chat messages pose particular challenges for keyword searches, as employees trade shorthand, slang, and references that make sense to one another but escape carefully drafted queries. Through Pso-Rite.com LLC v. Thrival LLC and Kim v. Cushman & Wakefield U.S., Inc., Favro showed how courts have required additional search efforts when keywords alone failed to account for how people actually communicate. Neither court prescribed a particular technology, but both made clear that parties needed to look beyond their initial searches to meet their discovery obligations. In his analysis, Favro considered how natural language queries using AI and concept-based clustering can uncover relevant exchanges, while recognizing that costs, proportionality, and the facts of each matter should guide the choice of methods. Read the article below to learn where search terms still serve a useful role and when the evidence calls for a more thoughtful approach to text and chat messages.
Courts Highlight Limitations with Text and Chat Message Searches
By Phil Favro, Contributing Author for HaystackID
With the advent of artificial intelligence (AI), commentators have suggested that search terms will finally phase out from being used for eDiscovery search.
Not exactly.
Search terms continue to be used and with frequency. This is confirmed by anecdotal reports, along with actual cases. In fact, it appears that many parties continue to use search terms as their primary methodology for identifying relevant information. In addition, parties—rather than dispensing with search terms entirely—are using them in connection with AI-assisted review workflows. This is apparent from the recent Schulte v. LinkedIn Corporation case, where a court approved the defendant’s use of search terms to pre-cull irrelevant information from a set of documents before using its AI-assisted review technology to identify relevant information for production. As Schulte explained, using search terms in this context is both reasonable and proportional, particularly since a producing party would incur “significant costs related to processing, hosting, and human review” if it were to run the entire document set through the AI technology.
If Schulte is any indicator, parties may continue using search terms with AI, just as they have done with technology-assisted review (TAR) workflows. In other words, for eDiscovery workflows, search terms (for the time being) might be here to stay.
Limited Utility of Search Terms
And yet, there are other contexts where search terms may not be viable at all. To be sure, search terms have always had their limitations. Courts and cognoscenti have emphasized for years the problems of over-inclusiveness (including far too many documents for review that have nothing to do with the particular query), together with under-inclusiveness (failing to pinpoint many relevant documents because they do not include any of the words comprising the search terms).[1]
Parties can ameliorate those problems in certain instances through testing, sampling, and other quality control measures. But even the most fastidious quality assurance process may not be enough to make search terms usable in some contexts without the help of other methodologies or analytics tools. One of those areas involves searches of text messages and messages exchanged on chat and collaboration tools like Google Chat, Slack, and Microsoft Teams.[2]
Searches of Text Messages and Chat Messages
Text messages and chat messages are often key forms of evidence in litigation. Regarding text messages, the Honorable Richard Sullivan of the U.S. Court of Appeals for the Second Circuit recently explained that “private statements in text messages may reflect a party’s unfiltered views . . . [and] a different perspective on his state of mind than those he made publicly.”[3] This evaluation of the importance of text messages has been confirmed on multiple occasions, with courts and others opining that “text messages generally memorialize unguarded communications and contemporaneous observations that cannot be replicated through other forms of discovery.”[4] Those observations are often applicable to chat messages.
What makes these communications so valuable in terms of evidence—their informality and the tendency of those communicating to relax and engage in informal, unscripted dialogue—also renders them difficult to explore using search terms. The use of slang and other informalities creates difficulty in crafting and refining terms that can effectively target relevant information.
Lawyers and technologists are not the only ones aware of this phenomenon. Courts have caught on to this issue and ordered parties to consider additional search strategies for identifying relevant materials among text messages and chat messages. Two recent cases that considered this issue are Pso-Rite.com LLC v. Thrival LLC and Kim v. Cushman & Wakefield U.S., Inc. Both Pso-Rite and Kim highlight the limitations of using search terms to find relevant information in text messages and chat messages. They also emphasize the need to consider other methodologies and tools, including AI technologies, to facilitate this process.
Pso-Rite – Text Messages and Chat Messages
In Pso-Rite, which involved claims and counterclaims arising from the marketing and sale of massage products, the parties had several disputes regarding the plaintiff’s production of documents, including text messages.[5] The plaintiff’s text message production format was PDF, predominantly featuring screenshots of relevant text messages. The text message production did not include metadata or sender, recipient, or date and time stamp information.
The court-appointed special master concluded that the plaintiff’s text message production was not reasonably usable under Federal Rule of Civil Procedure 34(b)(2)(E)(ii) and ordered the plaintiff to redo its production. In addition, to ameliorate concerns that the production itself was under-inclusive of relevant text messages, the special master also directed the plaintiff to image the phones from two key custodians and then conduct searches for relevant text messages and chat messages. While acknowledging the parties’ collective efforts to develop search terms to facilitate this process, the special master also determined that the plaintiff needed to undertake further efforts to locate responsive information:
Pso-Rite must do more than just use search terms to locate relevant text messages since search terms may not hit on responsive information given that texts are replete with idioms, slang, jargon, and other contextual language. (emphasis added)
The special master made clear that manually reviewing text messages and chat messages was not required, nor would he require the plaintiff to select a particular methodology to enhance its search efforts. Nevertheless, the plaintiff had to do more than just use search terms given their inherent limitations and the plaintiff’s obligation under Rule 26(g)(1) to conduct a good faith, reasonable inquiry for relevant information in response to the defendants’ written discovery responses.
Kim – Chat Messages
Similarly, the court in Kim found that searches through chat messages with search terms would likely be ineffective.[6] Kim involved claims arising from alleged pregnancy discrimination, and the court (Magistrate Judge Steve Kim) reasoned that chat messages from the corporate defendant’s Microsoft Teams environment could be particularly relevant given that it was one of the defendant’s principal communication tools for its employees.
Nevertheless, the defendant—in connection with its production efforts—inadvertently neglected to search through these communications. Upon discovering this omission, the defendant ran search terms across the Teams messages it collected, producing 47 pages of responsive messages. Despite this production, Judge Kim observed that the defendant’s search terms did not identify two messages discussing the plaintiff’s efforts to transition her job duties before her pregnancy leave. Reflecting on the limitations of using search terms to conduct this search, Judge Kim indicated that “[t]hese sorts of communications are relevant to the pretext analysis in this case.”
As a result, the court ordered the defendant to undertake an additional search of its Teams messages and make a supplemental production of responsive documents. In so doing, Judge Kim dismissed the defendant’s argument that it “already ran the search terms against Teams and there’s nothing left.” The judge reasoned that the terms were too under-inclusive as they were all tethered to the plaintiff’s full name (Connie Kim); for example, “Connie Kim” NEAR “terminat!”. While those terms may or may not identify responsive information among emails, the court was nearly certain they would be inadequate for locating relevant details among Teams messages:
It is arguable whether that may work well enough even for emails, but it cannot work for MS Teams chats about transition planning among managers who might say “the Smartsheet” or “Brooke’s workload” without mentioning Plaintiff by name. Keyword searches alone, without more advanced and thoughtful search techniques, will be inadequate for Teams data—a medium where conversations are shorter, more informal, and less likely to include full names than email.
Like the special master in Pso-Rite, Judge Kim ultimately declined to order the defendant to use a specific search methodology. Instead, the defendant was “free to consult with its eDiscovery consultant on the most efficient and defensible methods for searching, reviewing, and producing Teams data.” The court provided a framework for those searches by ordering the defendant to search the Teams messages of three custodians during a six-month period, with specific guidance on the nature of the messages to be produced in discovery.[7]
Additional Methods for Searching Text Messages and Chat Messages
Both Pso-Rite and Kim spotlight the limitations of search terms for identifying relevant information among text messages and chat messages. Pso-Rite emphasized that search terms may not be effective given the presence of “idioms, slang, jargon, and other contextual language.” Kim made clear that search terms were deficient given that chat messages are “shorter, more informal, and less likely to include full names.” While search terms could yield some responsive information, they will probably be inadequate “without more advanced and thoughtful search techniques.”
In lieu of search terms, what are some “advanced and thoughtful search techniques” for handling searches of these communications?
Parties can use natural language searches using AI technology. Parties whose eDiscovery platforms incorporate AI could run natural language queries to find documents that search terms might otherwise leave undetected. Unlike search terms, AI searches penetrate beyond a document’s words to potentially glean context, meaning, and intent. For example, for the pregnancy discrimination claims in Kim, the defendant could use its AI-powered eDiscovery technology to query for messages reflecting pregnancy transition discussions. Depending on the functionality and features of the application, the defendant may be able to pinpoint documents about the topics at issue that might remain hidden from search terms.
Clustering also stands out as a more effective method than just search terms alone. With clustering technology, parties can analyze communications grouped by concept or topic. Clustering can also be used with search terms to more effectively find common language themes that parties may then translate into search terms that target relevant information with greater precision.
Whether a party selects AI, data clustering, or other search methods or analytics tools for handling the identification of relevant text messages and chat messages will depend on the facts and circumstances of a particular matter. Costs and other factors will influence the selection of eDiscovery search tools. In low-value cases, search terms may well represent a reasonable and proportional solution. And yet, in matters where text messages and chat messages are key evidence, Pso-Rite and Kim make clear that parties should consider additional search methods to enhance their retrieval efforts for this source of relevant information.
[1] Deal Genius, LLC v. O2COOL, LLC, 682 F. Supp. 3d 727, 734 (N.D. Ill. 2023). [2] During this article, the term “text messages” will refer to messages exchanged on text messaging applications such as iMessage and Google Messages; and instant, encrypted, or ephemeral messaging applications like WhatsApp and Signal. Messages exchanged on chat and collaboration tools will be collectively referred to throughout this article as “chat messages.” [3] Oakley v. MSG Networks, Inc., 792 F. Supp. 3d 377, 388 (S.D.N.Y. 2025). [4] Lee v. Cnty. of Los Angeles, No. 2:23-CV-06875-GW-MAA, 2025 WL 2505484, at *20 (C.D. Cal. Aug. 29, 2025), reconsideration denied, No. CV 23-6875-GW-MAAX, 2025 WL 3211209 (C.D. Cal. Oct. 14, 2025). [5] Pso-Rite.com LLC v. Thrival LLC, No. 1:21-CV-00775-PAB-STV, 2025 WL 3899841 (D. Colo. Dec. 23, 2025). The author served as the court-appointed special master in this litigation and was responsible for the issuance of the order discussed in this article. [6] Kim v. Cushman & Wakefield U.S., Inc., No. 2:25-CV-4783-CAS (SKX), 2026 WL 1353455 (C.D. Cal. Apr. 24, 2026). [7] The court provided these details in order to circumscribe the nature and extent of the search given the plaintiff’s request “for a nearly indiscriminate search of ‘all reasonably likely repositories.’”
About Phil Favro
Phil Favro is the founder of Favro Law PLLC, where he counsels clients on ESI, AI, and discovery issues and serves as a special master, mediator, and expert witness. Phil is nationally recognized for his expertise on ESI, discovery, and information governance, with courts acknowledging his credentials. See, e.g., Oakley v. MSG Networks, Inc., No. 17-CV-6903 (RJS), 2025 WL 2061665 (S.D.N.Y. July 23, 2025). This background makes Phil particularly well-suited to counsel clients and advise courts on information-related issues. As a special master, Phil is acclaimed for his collaborative approach, working with parties to find stipulated solutions to complex issues. For disputes that require adjudication, he is renowned for the clarity and vigor of his written dispositions, which are available on legal search engines.
About HaystackID®
HaystackID® solves complex data challenges related to legal, compliance, regulatory, and cyber requirements. Core offerings include Global Advisory, Cybersecurity, Core Intelligence AI™, and ReviewRight® Global Managed Review, supported by its unified CoreFlex™ service interface and eDiscovery AI® technology. Recognized globally by industry leaders, including Chambers, Gartner, IDC, and Legaltech News, HaystackID helps corporations and legal practices manage data gravity, where information demands action, and workflow gravity, where critical requirements demand coordinated expertise, delivering innovative solutions with a continual focus on security, privacy, and integrity. Learn more at HaystackID.com.
Assisted by GAI and LLM technologies.
SOURCE: HaystackID