Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Ensure the future of local Bay Area News by becoming a Local News Matters member today.
Thanks for your contribution!
Receive in-depth news stories and arts & entertainment coverage from around the Bay Area in your inbox.
Local News Matters
Connecting audiences with quality, local news
A UTAH MAN IDENTIFIED ONLY AS JOHN DOE brought a class action lawsuit last week against Perplexity, the San Francisco based AI developer that contends the company embedded Meta and Google tracking technologies in its code. Meta and Google were also named as defendants.
Ahmad, Zavitsanos & Mensing PLLC, the Houston-based law firm that brought the case on behalf of an unidentified “John Doe” plaintiff, filed a 135-page complaint that lays out Doe’s story in significant detail.
Don’t miss out on Bay Area news, delivered to your inbox twice a week.
Doe’s suit contends that the tracking technology intercepted and sent to Meta and Google the text of the users’ questions and Perplexity’s responses, all without disclosure to those users. The complaint alleges that users did not consent to share what could be highly personal matters — private questions about legal, political, financial, sexual and health issues — with third parties, particularly two monoliths whose business plans involve collecting personal information for use in targeting advertisements.
“Perplexity effectively planted a bug on plaintiff’s and class members’ web browsers and compelled them to unknowingly disclose their confidential communications to Meta and Google,” according to the complaint.
The centerpiece allegation in the complaint is that the sharing happened even if the user posed their searches in “incognito mode” and expected the data to be kept strictly confidential.
No response to the 135-page lawsuit has yet been filed by Perplexity. At this stage, the allegations in the complaint are unproven.
No response to the 135-page lawsuit has yet been filed by Perplexity. At this stage, the allegations in the complaint are unproven.
The complaint asks the federal court to certify a nationwide class of Perplexity users (excluding paid “Pro” and “Max” subscribers) who asked questions that were tracked during a four-year period that ran through February 2026.
Perplexity is alleged to have a $20 billion valuation, making it a tiny player compared to two other San Francisco AI developers — Anthropic and OpenAI — but a $20 billion company is still an attractive defendant for a class action claim.
No response to the lawsuit has yet been filed by Perplexity, and at this stage, the allegations in the complaint are unproven.
While a search of recent litigation did not identify any other suits against Perplexity or other major AI developers based on the plaintiffs’ tracking allegations, there has been an explosion in litigation against tracking technologies used in other contexts, and many such cases — like Doe’s — have been filed in federal court in San Francisco.
Those lawsuits have challenged undisclosed tracking technologies used on websites and apps of hospitals, medical professionals and banking organizations, among others. Many of the cases assert claims under federal and California statutes that prohibit individuals from surreptitiously intercepting communications, the practice generically referred to as “wiretapping.” While the statutes that forbid the practice focus on telephone calls, plaintiffs argue that intercepting communications transmitted over the internet is also covered. The courts have not finally resolved all the legal questions, but at least one decision supports the plaintiffs’ theory.
Chatting with the chatbots
Bay City News recently wrote a story about Doe’s lawsuit and contacted Perplexity AI, Inc. (the Perplexity chatbot’s owner) with a request for comment on the case. A company spokesperson responded promptly to say they couldn’t comment as the plaintiff had yet to formally “serve” (or deliver) the suit.
Given our coverage of the lawsuit and the fact that AI platforms regularly provide analyses of lawsuits and litigation, we thought it would be of interest to see how Perplexity — not the company, but its AI chatbot — evaluated the lawsuit. To that end, we uploaded the plaintiffs’ 135-page complaint to the Perplexity platform and asked the bot what it thought about the suit. We also asked Claude and ChatGPT (the chatbots of Anthropic and OpenAI respectively) about the suit against Perplexity. (Their reactions are in the sidebar following this story.)
Perplexity weighs in
Support our independent, nonprofit newsroom, Local News Matters, by becoming a member today. Members enable us to pay reporters, photographers and editors to serve our communities with local news that matters in the greater Bay Area.
Perplexity has an easy and friendly conversational style that imitates human interaction, though more formal than typical human texting. Despite that, Perplexity did not directly acknowledge that it — Perplexity — was the AI discussed in the lawsuit, and it referred to Perplexity in the third person.
Before it began, Perplexity stated what it could do and what it could not — saying it could not give “jurisdiction specific legal advice” or “fill in missing law or facts.” Rather it said it must rely on the complaint and any cases it cites, “because AI tools are known to hallucinate citations or misstate holdings if they try to improvise.”
It added, “I cannot replace proper legal research or a licensed lawyer’s judgment about viability, risk, or case strategy.”
With those cautions in place, Perplexity said, “the complaint tells a story that is facially coherent and tailored to current privacy-litigation trends, and most of the core claims look at least colorably plausible.”
Perplexity began by identifying the strengths of the lawsuit. It said, “the complaint is long but not incoherent” and is strongest on its “narrative of sensitive use and betrayal of privacy.” It particularly noted the focus on AI-chat as “intimate, quasi therapeutic or advisory, including health, mental health, finance, legal, and relationship issues. …” It directly acknowledged that in talking to an AI bot, “people over disclose precisely because they expect anonymity or at least confidentiality.”
Perplexity said that among the allegations, the claims about sharing data while in “incognito mode” were especially powerful.
On the negative side, it thought the defendants would push hard to prove there was user consent to data sharing through the app’s terms of service and banners on the site. It thought that if the AI shared full transcripts of user interactions it would be a strong point for plaintiff as the case progresses.
It also noted several other challenges for the plaintiff, including the difficulties in obtaining certification of a class, navigating the requirements of federal court jurisdiction, and proving harm was caused to class members.
Perplexity emphasized that it was reading a complaint — a set of allegations drafted by the plaintiff to put his claim in the best light — not “findings” made by a court or jury after hearing evidence. In response to the data-tracking allegation, Perplexity said, “the ‘embedded in the code’ story will rise or fall on what discovery shows.” Discovery refers to the legal process in which both sides in a case exchange documents and question witnesses under oath.
Settlement value
We asked Perplexity to assess the “settlement value” of Doe’s case.
Lawyers frequently talk about cases’ settlement value, or the predicted amount at which the case would settle. There is no universal definition for how settlement value is determined and the amounts change over the course of a case depending on the stage, the unearthed evidence, the quality of the lawyer and numerous other factors.
Settlement value is also very much in the eye of the beholder; plaintiffs, defendants and judges may also quantify the settlement value of a particular case in very different amounts as seen from their different vantage points.
Lawyers typically don’t like to be pinned down to a specific settlement value when talking to their clients for fear that developments in the case will prove them wrong.
Perplexity did not want to give numbers. “At this stage, the only honest answer is that settlement value is highly uncertain.”
We did not let Perplexity off the hook; we asked if it could give a range.
From other interactions, Perplexity knows that our BCN reporter is a journalist. When we pressed for concrete settlement values it said, “You can plausibly frame it for readers as ‘mideight figures is very much on the table if the case goes well for plaintiffs, but the realistic band is wide.’”
When we pressed further — and after some back and forth — Perplexity finally agreed that this statement is accurate and not misleading: “Based on recent tracking [of] tech and privacy settlements generally, rough settlement ranges for this case run from lower band $5-15 million, middle band $15-50 million to upper band $50-150 million, though the actual outcome in this case could land below, within, or above that band, depending on how the facts and rulings develop.”
Determining John Doe’s odds
The takeaway based on the bot’s analysis of the complaint is that John Doe is off to a good start, telling a story that generally seems grounded and coherent. If the facts alleged can be proven, particularly the allegations that the full text of user prompts and responses are tracked and shared, plaintiff is likely to gain traction as the case progresses.
Perplexity thinks that the sensitivity of user data and the allegedly misleading workings of the “incognito mode” are features that will aid plaintiff in pushing for a higher amount.
And the settlement value?
The bot was hard to pin down about a specific settlement value. In the end, it said the case could be in the $5 million to $150 million range, or it could be higher or lower.
Based on that answer, it is very unlikely that Perplexity will be proven wrong.
At the beginning of its analysis, Perplexity advised AI chatbots “cannot replace proper legal research or a licensed lawyer’s judgment about viability, risk, or case strategy.”
But based on its prediction of settlement value, human lawyers will welcome Perplexity to the club.
WHAT OTHER AI CHATBOTS SAY
To get perspective on how Perplexity — the chatbot — would evaluate John Doe’s class action against the company Perplexity, we uploaded Doe’s complaint to Anthropic’s Claude and OpenAI’s ChatGPT to get their perspectives.
➤ CLAUDE
Our reporter said to Claude, “Here is a new lawsuit. Would you analyze it and let me know what you think about the plausibility of the allegations and the legitimacy of its claims?”
Claude took about a minute to digest the 135 pages in the complaint. It began, “This is a serious, sophisticated privacy class action, and — at least on its face — it’s not frivolous.”
Claude pointed out that at this stage, the lawsuit contained unproven allegations and whether the plaintiff could prove them was unknown. Focusing on the allegations, Claude said they were “moderately strong” from a technical plausibility standpoint, but it honed in on the question of what data was actually disclosed to Google and Meta. The complaint said the data included user “prompts, follow-ups and URLs with content.” Claude thought those would be harder to prove than if the trackers were just picking metadata, clicks and events. It said that “if plaintiffs prove prompts were transmitted [and] URLs exposed full conversations, this becomes a high-risk case for defendants.”
Among the case’s strengths, Claude focused on the allegations that conversations were shared even when in incognito mode (“courts care a lot about this”) and that sensitive medical queries and financial discussions were allegedly shared (“judges react strongly”).
Claude expressed some skepticism that “complete transcripts” were shared with Google and Meta. (“That’s a very aggressive assertion and may be hard to prove.”) It also thought the plaintiff had lumped many different technologies together and the “defense will argue that the plaintiff misunderstands how these tools work.”
Claude then walked through the key stages of federal court litigation, offering its assessment of how well plaintiffs would do at each stage. It then offered its overall take, saying this is “a serious, well-crafted complaint that may overstate its technical claims — but is legally viable.”
It then added its opinion that “the entire case turns on a single factual question: Did Perplexity actually transmit the substance of user conversations to Meta/Google? If yes — major exposure; if no — case collapses quickly.”
We asked Claude its opinion of the settlement value of the case.
In Claude’s view, the factors that would lead to greater settlement value were if the actual prompts were shared (“HUGE multiple”) as well as the sensitivity of tracked matters and the Incognito Mode allegations (“courts hate misleading privacy features”).
Claude then assessed the factors that would push the settlement value down, including proof problems, possible user consent issues under the app’s terms of service and the challenges of proving an entitlement to class certification.
Claude then offered its assessment of settlement value depending on how strong the evidence of content sharing was.
If low: $5 million to $15 million; high $75 to $150 million.
In the middle case — which it called the most likely scenario — $20 to $60 million.
Finally, Claude said, “If I had to price this today: $30 million to $50 million expected settlement value.”
➤ CHATGPT
ChatGPT said, “The core factual allegations are highly plausible” for three reasons. First, the technology is real and well-documented; second, the conduct is not novel and has been proven in other contexts; and third, the plaintiff’s investigation “appears concrete” because it “suggests actual forensic investigation, not boilerplate allegations.”
Chat GPT added, “The ‘incognito mode’ allegation is particularly damaging if true.”
ChatGPT thought the legal claims were a mixed bag but that the wiretap approach under state and federal law were the best claims, because of plaintiffs’ success with them in other contexts.
Its overall assessment: “The complaint is well-constructed and legally serious… it reflects a genuine forensic investigation, applies established legal theories that have produced verdicts against defendants in analogous cases, and targets conduct that, if proven as alleged, clearly violates California and federal privacy law.”
At the low end, the value was $8 to $15 million and at the high end $75 to $150 million.
The most likely outcome, in its opinion, was the mid-range estimate of $25 to $50 million, almost the same as Claude’s assessment.
Our reporter said to Claude, “Here is a new lawsuit. Would you analyze it and let me know what you think about the plausibility of the allegations and the legitimacy of its claims?”
Claude took about a minute to digest the 135 pages in the complaint. It began, “This is a serious, sophisticated privacy class action, and — at least on its face — it’s not frivolous.”
Claude pointed out that at this stage, the lawsuit contained unproven allegations and whether the plaintiff could prove them was unknown. Focusing on the allegations, Claude said they were “moderately strong” from a technical plausibility standpoint, but it honed in on the question of what data was actually disclosed to Google and Meta. The complaint said the data included user “prompts, follow-ups and URLs with content.” Claude thought those would be harder to prove than if the trackers were just picking metadata, clicks and events. It said that “if plaintiffs prove prompts were transmitted [and] URLs exposed full conversations, this becomes a high-risk case for defendants.”
Among the case’s strengths, Claude focused on the allegations that conversations were shared even when in incognito mode (“courts care a lot about this”) and that sensitive medical queries and financial discussions were allegedly shared (“judges react strongly”).
Claude expressed some skepticism that “complete transcripts” were shared with Google and Meta. (“That’s a very aggressive assertion and may be hard to prove.”) It also thought the plaintiff had lumped many different technologies together and the “defense will argue that the plaintiff misunderstands how these tools work.”
Claude then walked through the key stages of federal court litigation, offering its assessment of how well plaintiffs would do at each stage. It then offered its overall take, saying this is “a serious, well-crafted complaint that may overstate its technical claims — but is legally viable.”
It then added its opinion that “the entire case turns on a single factual question: Did Perplexity actually transmit the substance of user conversations to Meta/Google? If yes — major exposure; if no — case collapses quickly.”
We asked Claude its opinion of the settlement value of the case.
In Claude’s view, the factors that would lead to greater settlement value were if the actual prompts were shared (“HUGE multiple”) as well as the sensitivity of tracked matters and the Incognito Mode allegations (“courts hate misleading privacy features”).
Claude then assessed the factors that would push the settlement value down, including proof problems, possible user consent issues under the app’s terms of service and the challenges of proving an entitlement to class certification.
Claude then offered its assessment of settlement value depending on how strong the evidence of content sharing was.
If low: $5 million to $15 million; high $75 to $150 million.
In the middle case — which it called the most likely scenario — $20 to $60 million.
Finally, Claude said, “If I had to price this today: $30 million to $50 million expected settlement value.”
➤ CHATGPT
ChatGPT said, “The core factual allegations are highly plausible” for three reasons. First, the technology is real and well-documented; second, the conduct is not novel and has been proven in other contexts; and third, the plaintiff’s investigation “appears concrete” because it “suggests actual forensic investigation, not boilerplate allegations.”
Chat GPT added, “The ‘incognito mode’ allegation is particularly damaging if true.”
ChatGPT thought the legal claims were a mixed bag but that the wiretap approach under state and federal law were the best claims, because of plaintiffs’ success with them in other contexts.
Its overall assessment: “The complaint is well-constructed and legally serious… it reflects a genuine forensic investigation, applies established legal theories that have produced verdicts against defendants in analogous cases, and targets conduct that, if proven as alleged, clearly violates California and federal privacy law.”
At the low end, the value was $8 to $15 million and at the high end $75 to $150 million.
The most likely outcome, in its opinion, was the mid-range estimate of $25 to $50 million, almost the same as Claude’s assessment.
Joe Dworetzky is a second career journalist. He practiced law in Philadelphia for more than 35 years, representing private and governmental clients in commercial litigation and insolvency proceedings. Joe served as City Solicitor for the City of Philadelphia under Mayor Ed Rendell and from 2009 to 2013 was one of five members of the Philadelphia School Reform Commission with responsibility for managing the city’s 250 public schools. He moved to San Francisco in 2011 and began writing fiction and pursuing a lifelong interest in editorial cartooning. Joe earned a Master’s in Journalism from Stanford University in 2020. He covers Legal Affairs and writes long form Investigative stories. His occasional cartooning can be seen in Bay Area Sketchbook. Joe encourages readers to email him story ideas and leads at joe.dworetzky@baycitynews.com.
Bay City News Foundation
(510) 251-8100
newsroom@baycitynews.com
Staff Page
Terms of Use
Sign in by entering the code we sent to , or clicking the magic link in the email.
Terms & Conditions. This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.