Google AI Overview lawsuit dismissed: what AEO teams do next

2 October 2026 · The Cresia team

It means publishers can't count on a court to force Google to pay for, or send traffic in return for, content that appears in AI Overviews, at least not on the theory Penske tried. Search Engine Roundtable reports the case was dismissed because Penske couldn't show that Google had entered any formal agreement to trade its content for some amount of traffic. For AEO and GEO teams the working answer is simple: visibility in AI answers is something you earn and measure, not something you're owed.

Key takeaways

The Penske case was dismissed because no formal agreement with Google was shown, so there is no legal shortcut to traffic from AI Overviews.
Visibility in AI answers is earned passage by passage. Plan around that, not around lawsuits.
Make key claims easy to lift: one question per section, the answer first, named entities, dated facts.
Measure citations, brand mentions and referral sessions across a fixed query set, and treat any single AI answer as unstable.
Don't block crawlers or rewrite your site because of one ruling. Decide per content type what you want shown, and check each vendor's own documentation for the controls.

What the Penske dismissal actually says

Penske Media owns Rolling Stone, The Hollywood Reporter, Billboard, Variety and other titles. According to Search Engine Roundtable, it sued Google about a year ago over AI Overviews. The case has now been dismissed, and the stated reason is narrow: Penske did not prove that Google had agreed to anything formal, such as selling its content in exchange for a given amount of traffic.

Read that carefully. It isn't a finding that AI Overviews are good for publishers. It isn't a finding that they're harmful. It says that, on the facts presented, the claimed bargain didn't exist on paper.

That distinction matters because the industry's working assumption for twenty years was an informal one: you let search engines crawl you, they send visitors. Nobody signed it. The dismissal is a reminder that an assumed deal is not an enforceable one, and that the assumption gets weaker as answers move onto the results page itself.

What we can't say is what happens next. Appeals, refiled claims and other cases may go differently, and the details of the ruling are best read from the reporting and the court record rather than summarised from a headline. Your plans shouldn't depend on any of it.

Why this matters for AEO and GEO programmes

Most AEO and GEO work quietly rests on a belief that if the content is good and accessible, the engine will cite it and some traffic will follow. The dismissal doesn't break that belief, but it removes one backstop. There is no contractual floor. If a Google answer summarises your article and nobody clicks, that outcome is currently a product decision, not a breach of anything.

That changes how you should talk about the work internally. A head of SEO who promises leadership that AI search visibility will recover lost clicks is making a promise the platform hasn't made. A better framing: AI answers are another surface where your brand can be named, quoted or linked, with uncertain referral value, and the job is to be present accurately.

Two groups feel this differently. Large publishers with licensing leverage can negotiate directly with AI companies, and that's a business-development question, not an SEO one. Everyone else, which is nearly every in-house team reading this, has no such lever. For them the only lever is the quality and structure of what's on the page.

There's also a quieter point. Search Engine Roundtable has been covering AI Overviews returning for many large brand-name queries. Branded queries are where teams used to feel safe. If answers now appear there too, defending your own name in the answer box becomes a real task, not a theoretical one.

How AI answer engines choose and cite sources

No vendor publishes its full selection logic, and it changes. But the general mechanics are stable enough to plan around, and each one has a practical implication.

Retrieval first. The system fetches candidate pages for the query, and often for several related sub-questions it generates itself. If you only target the head term, you're missing the sub-questions where citations are decided.
Passage-level extraction. The engine lifts a paragraph or a table row, not your whole page. A passage that answers its own heading in two or three sentences is easier to use than one that depends on the three paragraphs before it.
Entity and consistency checks. Names, products, prices and dates that agree across your pages, your structured data and third-party mentions give the engine less reason to distrust you.
Access and eligibility. If a crawler can't fetch the page, or you've told it not to show a snippet, you're out. Controls differ by vendor, and the one that governs a Google search feature isn't necessarily the one that governs a chatbot's training. Check Google's and each vendor's documentation before you edit robots rules.
Source diversity. Answers often draw on forums, reviews, video and reference sites alongside publishers. Your own domain is one input among several.

Structured data sits in that picture as clarification, not a cheat code. Search Engine Journal has covered common structured data mistakes that hurt AI visibility, and the theme is mundane: markup that contradicts the visible page, stale values, and types applied to content that doesn't match. Fix those before you chase anything exotic.

What to change in your content this quarter

Start with pages that already earn impressions on informational queries. Rewriting the whole library is a trap; a dozen well-chosen pages will teach you more.

A worked example. Say you run SEO for a payroll software company, and your page on 'how to calculate overtime' opens with a story about how work has changed. An engine looking for a passage that defines the calculation will skip it. Move the definition and one worked calculation to the top, put the exceptions under their own heading, and date the page when the rules last changed. Same page, same author, far easier to quote.

Concrete steps:

Put the direct answer in the first two sentences under each question-style heading.
Give each section one job. If it answers two questions, split it.
Replace vague claims with named, dated, attributable ones, or cut them.
Keep entity names consistent: product names, people, locations, the way you write your own brand.
Make sure each author and reviewer shown on a page is a real, consistent record, not a different byline on every template.
Check that visible content and structured data say the same thing.
Add comparison tables where the reader is choosing between options. Tables are easy to extract and hard to misread.

And one that isn't about words: make sure the page loads the answer in the HTML, not behind a script that a crawler may never run.

How to measure AI search visibility without fooling yourself

Answers vary by query phrasing, location, session and day. A single screenshot proves nothing; a week of screenshots proves little more. You need a fixed query set, a regular sampling rhythm and a record of what you saw.

Build the set from three buckets: branded queries, category queries where you should appear, and comparison queries where a competitor is named. Run the same set on the same schedule across the engines your buyers use. Log whether you were cited with a link, mentioned without a link, or absent, and what the answer said about you.

A platform built for this can save time; MediaPilot is Cresia's product for media and AI-search visibility, and the MediaPilot page shows what it covers. Spreadsheets work too if the sampling is disciplined.

Measure	What you record	Owner	Cadence
Citation presence	Cited with link, mentioned only, or absent, per query	SEO lead	Weekly
Answer accuracy	Whether claims about your product or brand are correct	Content lead	Weekly
Referral sessions	Visits from AI surfaces, segmented from organic search	Analytics lead	Monthly
Branded query share	Whether an AI answer now sits above your own listing	SEO lead	Monthly
Source mix	Which domains the engine cites alongside or instead of you	PR / comms	Monthly

Referral measurement is the weak spot. Some AI surfaces pass identifiable referrers and some don't, so a flat line in analytics may be a tagging gap, not an absence of visits. A clean tracking specification helps you tell the difference; the explainer on tracking specifications covers the idea, and OmniSpec is the product for analytics governance.

What not to do after this ruling

The temptation after a court story is to make a sweeping move. Resist most of them.

Don't block every AI crawler in a panic. Blocking may be right for some content, such as paid research or licensed material. For public marketing pages, it usually just removes you from answers your buyers are reading.
Don't assume a robots.txt line controls everything. Different user agents govern different products. Verify against the vendor's documentation, then test.
Don't chase one answer. If a single AI Overview omits you today, that's an observation, not a diagnosis. Sample it again.
Don't write for the engine at the expense of the reader. Passages built purely to be quoted, with no argument or evidence, read as filler and tend to age badly.
Don't stuff schema. Markup that describes things the page doesn't contain is a liability.
Don't promise traffic. Report citations and mentions as visibility, and report sessions only when you can attribute them.

There's a version of this advice that says do nothing until the legal picture clears. That's wrong too. The mechanics of being cited don't depend on any lawsuit, and the pages you fix now are better for human readers regardless.

Deciding what you want shown

The more useful question this ruling raises is internal: which of your content do you want an engine to summarise, and which do you want people to visit?

Split your library roughly in two. Content whose job is awareness, such as definitions, how-tos and category explainers, gains from being quoted accurately even if the click never comes, because the brand name travels with it. Content whose job is conversion or depth, such as calculators, tools, gated research and pricing logic, loses value when flattened into a summary.

For the second group, make the page worth the visit: the interactive element, the full dataset, the downloadable template. Give the summary a reason to point at you instead of replacing you. For the first group, make the extractable passage excellent and accurate.

This is also where media and content teams need to talk. Paid and organic teams both see AI answers eating the top of the page; the media teams solution page frames how that conversation fits into planning. Whoever owns the brief should write down, per content type, what outcome counts as success.

Frequently asked questions

Does the dismissal mean Google can use any publisher's content in AI Overviews?

Not as a general legal conclusion. Search Engine Roundtable reports the dismissal turned on Penske not proving a formal agreement with Google about content and traffic. Other claims, other plaintiffs and other courts may land differently, so treat it as one data point, not a settled rule.

Should we block Google's crawlers from our content now?

Usually not for public marketing and educational pages, because blocking removes you from the answers your buyers read. Do it selectively for content you license or sell, and confirm in Google's documentation which control governs which feature before changing anything.

What's the difference between AEO and GEO?

In practice the two overlap heavily. AEO usually refers to being the answer in search features and assistants, and GEO to being cited inside generated responses. The work is the same core discipline: clear, extractable, accurate passages on pages that engines can fetch.

How do we report AI search results to leadership without overselling?

Report presence and accuracy first: how often you're cited or mentioned across a fixed query set, and whether what's said about you is right. Add referral sessions only where you can attribute them, and say plainly where tracking has gaps.

Sources

https://www.seroundtable.com/google-ai-overview-lawsuit-dismissed-42211.html
https://www.searchenginejournal.com/what-are-common-structured-data-mistakes-that-hurt-ai-visibility-ask-an-seo/589924/
https://www.seroundtable.com/google-ai-overviews-large-brand-names-42195.html