Trial Update, October 3, 2025: On rebuttals day, Google again desperately tries (and fails) to silence Stephanie Layser

Day 10 of US v. Google adtech remedies begins with Judge Brinkema welcoming a law school class. She jests that they are studying antitrust, and watching the trial today may turn them off, before taking judicial notice of all the trees that have been sacrificed in the course of the proceedings to create the massive binders of evidence and deposition transcripts accompanying each witness.

Stephanie Layser, Worldwide Head of Publisher AdTech Solutions, Amazon Web Services

Stephanie Layser, formerly of NewsCorp takes the stand for remedies rebuttals, after her powerful testimony during liabilities trial. Google has fought hard to keep her off the stand, absurdly questioning her firsthand knowledge of the publisher perspective in their motion to exclude. The objection-fest begins immediately, as DOJ begins their direct examination. Layser explains that she currently helps AWS publisher customers with their deployment of prebid servers, that she has consulted for small publishers, worked publisher-side for 12 years, and sat on the Prebid board.

Judge Brinkema sustains a couple of objections before seemingly realizing that Google simply does not want her to be heard. Among those sustained, Layser attempted to explain that DfP isn’t free to small publishers - they pay with their data, and that the supracompetitive AdX take rate of 20% is substantially higher than the pennies paid for ad serving. Judge Brinkema notes that this doesn’t answer the question of what would stop a DfP buyer from raising fees. She says, “I could almost answer….we don’t know who the buyer would be, what they would do, or what the market would look like,” before moving the examination along.

DOJ asks Layser to respond to the statement that “without Google to shepherd display advertising, monetization would stop and the internet could fall away.” Google’s Justina Sessions objects, and Judge Brinkema says that the statement is hyperbolic. DOJ’s David Geiger reminds the Court that this hyperbole was verbatim from WikiHow CEO Elizabeth Douglas’ testimony on Wednesday, and is permitted to proceed. Layser says that Google has controlled 90% of the ad serving market, and we have seen the effects it has had. Lower revenue for publishers. Lack of competition in the exchange market. Innovation ceased to exist, and publisher feature requests went unfulfilled. She brings up the example she used during liabilities of NewsCorp’s attempt to get access to real-time bid data from the ad server to help inform their personalization of their audience experience and optimization of NewsCorp’s monetization strategy, e.g. as between ads and subscriptions. Google withholding this data stalled publishers then, and stalls publishers now.

Layser is asked about her reaction to the statement that if Google divested AdX, Google’s demand would move to O&O. She says, “They already do that,” explaining that Google has already been prioritizing its owned properties. Layser explains why Google’s proposal for AdX to bid into rival ad servers would not eliminate switching costs. Publishers need the decision logic, and their log-level data to see direct against programmatic, evaluate incremental revenue value, and make informed decisions.  Similarly, she explains why the Google’s data sharing proposals don’t solve her concern, as this data would omit preferred deals and private auction deals, and the historical data could well be aggregate, and not the impression-level data publishers would need. DOJ’s proposal would afford publishers the same access to their own data that Google itself has.  We hear about NewsCorp’s Project Cinderella, where without the data, Layser couldn’t quantify the risk to revenue of switching ad servers, and had to assume all Google Ads demand would be lost, leading NewsCorp to abandon the effort to switch. When they went back to the table to negotiate DfP terms with Google, Google rejected nearly all of their redlines. Without power, and without other competitors in the market, NewsCorp had no choice but to stay.

She says she is “scared” of the language in Google’s proposal that defines “Qualifying Open Web Display Inventory” to mean Open Web Display Inventory offered for sale…by a Google Ad Manager Publisher customer in good standing and otherwise complying with the terms and conditions of the Publisher’s relevant contractual agreement(s) with Google.” Clearly, publishers have little say in the terms of these contractual agreements, and Google gets to decide who is in good standing. Sessions objects, and the Court asks if NewsCorp had been deemed by Google not to be in good standing. Layser explains the retaliation she faced from Google, being blacklisted from industry events. The Court notes that technically that is not the same as being found not in good standing, but I’m glad Layser managed to note this as the testimony is important in the context of DOJ’s non-retaliation provisions. Judge Brinkema asks if what it comes down to, like other witnesses, is that “you don’t trust Google?” Layser says yes.

Judge Brinkema poses a few telling hypotheticals. I am paraphrasing, though I tried to capture it to the best of my ability. If we have AdX out of the system, and if DfP stays in place but cannot treat AdWords differently. If there is an even playing field, and AdWords cannot preference DfP, what is the problem? Why is that not sufficient? “I worry there is something I haven’t thought of yet,” Layser explains, walking through the “moving targets” Google would set, and new conducts Google would launch. Detecting self-preferencing is very difficult without access to log-level data. Layser explains her past experience. “I knew something was weird,” but without access to the data, she felt “like a conspiracy theorist.” “It wasn’t until discovery” when all of Google’s auction manipulation started coming out of the woodwork that she realized that “she is not going crazy.” She felt validated. She says that’s why publishers need the data. “We need publishers to validate, and need the industry to be able to innovate on top of that so we can move forward from this.”

DOJ asks why it was important to her to be here today to testify. Layser explains that she cares deeply about the open web, and the business of publishing. She says that news, journalism, and free press is underpinned by this, and she wants to do anything she can to help.

Cross-examination was a barrage of failed attempts to undermine Layser’s credibility, as Sessions took advantage of Judge Brinkema’s waning patience to elicit yes or no answers to loaded questions and out-of-context statements, painting her as a Google critic. Judge Brinkema noted at one point that bias or prejudice is something that the finder of fact takes into consideration. Sessions brought up her recent “Scotch and Watch” podcast, and social media posts, trying at points to attack her character. It was poor form, particularly to the very witness Google employees had called “emotional and unproductive” for voicing the very concerns that the Court validated in its liability findings. DOJ, of course, allowed Layser to fill in the blanks on redirect, and her testimony was incredibly strong once again.

‍

Jason Nieh, Technical Feasibility Expert

This was about three hours of my life I will never get back. It was very difficult to parse how Nieh’s cross-examination landed, as he was incredibly smug and evasive, but Judge Brinkema, in the interest of moving things along, curbed a number of attempts by DOJ to expose his contradictions. DOJ began by highlighting that Nieh is not an adtech expert, never did a software migration at a big tech company, or replaced dependencies of software run by big tech companies.  What he lacks in practical experience, he makes up for as an expert witness, having served as a paid expert 28 times in 20 years. He served as an expert three other times for Google - the European Commission matter, Texas adtech antitrust case, and Search remedies. He also worked as an expert four times for Meta, and ten times for Apple.  In one of the Meta cases, the Judge found he was not qualified to render one of his opinions.

DOJ pointed out that Nieh misrepresented the assumptions DOJ’s experts relied upon in their analyses when he stated that neither expert accounted for functionality of divested products not being degraded, or limiting disruption to customers. DOJ asked if Nieh assumed that divestiture would not require Google to sell proprietary infrastructure. He did, at Google counsel’s instruction.  DOJ asked if Google providing a copy of some of these services would make divestiture go faster, and he qualified his answer, prompting Judge Brinkema to ask for a clear answer. He says yes. He coincidentally determined that divestiture would take no less than 5 years in his Search remedies work, just as he did here. He had no clear methodology apart from looking at cherry-picked “case studies” to support that determination.

He stuck to Google’s party line that the 2024 analysis contemplated a “business divestiture,” and not divestiture of a “functioning product,” and weaved around DOJ’s questions about where the customers would go as outlined in Google’s timeline if not to a working product. He described that this was a divestiture of things like staff and IP, not product. Yet, Heather Adkins had confirmed that source code is IP.

There was a lot he “didn’t recall” or hadn’t seen, despite testifying that he could have had access to whatever documents he deemed necessary from Google. He was not aware that OpenX - the largest independent exchange - migrated to GCP in seven months, for example. He didn’t know Google helped OpenX identify choose the right services on GCP to support the high volume, low latency needs they had. He didn’t find Google’s own internal divestiture feasibility analysis relevant to his assignment. He tried to avoid answering whether he surveyed suitable replacement candidates for dependencies on Google’s infrastructure, but Judge Brinkema prompted him to respond, where he admitted he did not. He didn’t recall Google engineer Sam Greenfield’s deposition statements about he did not do research to identify suitable replacements, though he relied on Greenfield’s declarations that there were none.  He never heard of Google’s ClangMR that enables “refactoring of large C++ codebases efficiently.”

There was a bunch of time wasted on redirect on whether a hypothetical code snippet presented by DOJ would “run.” DOJ’s point was that the code for a function only had to be modified once, and then the references to that function would need to be replaced, but Nieh had pretended not to understand the question any way it was worded. On redirect, he says he has no evidence it would run as written. Judge Brinkema suggests he wouldn’t know either way. He insists he knows it would not. DOJ came back to it on re-cross, and again it went nowhere.

What’s to come?

Trial will end Monday by 1:00PM. DOJ will call Matthew Wheatland, and possibly one of their technical experts. Google may call a surrebuttal witness to respond to Layser.

‍

‍