Quick Read
- Penske Media Corp sued Google, claiming AI Overviews drain traffic and revenue tied to search.
- Publishers are split: some litigate (e.g., against OpenAI and Microsoft), others sign licenses worth $1-$250m+.
- Core legal issues: copyright, compensation for training and summaries, and the sustainability of journalism.
- Control is limited: blocking AI summaries often means losing standard search visibility.
Why this matters for legal teams
Google's Enterprise Message Access and AI Overviews summarize answers inside the results page. That changes referral dynamics and the leverage publishers have in negotiations.
The legal exposure sits at the intersection of copyright, competition, and contract. Your organization's response will determine traffic, revenue, and bargaining power for the next several years.
The first big test case: Penske Media vs. Google
Penske alleges that about 20% of Google results that involve its content now surface AI Overviews, with clickthroughs and affiliate revenue falling sharply-more than one third lower than late 2024 levels. The complaint argues publishers cannot block inclusion in AI summaries without sacrificing conventional search placement.
Google's position: AI Overviews make search more useful, generate more diverse traffic, and still send billions of clicks daily. Expect Google to fight the case hard and resist any precedent that separates search indexing from AI summarization privileges.
What in-house counsel should track now
- Share of queries where your pages trigger AI Overviews vs. classic snippets.
- Clickthrough rates, session depth, and affiliate conversions before/after AO rollout.
- Search terms where AO fully answers a query (low intent to click) vs. partial answers.
- Any instances of factual lift, paraphrase that mirrors style, or quotation without attribution.
Publishers' pushback: lawsuits, licenses, and a split strategy
OpenAI and Microsoft face suits from major and regional outlets alleging unlicensed copying, memorization, and reproduction of articles. Several cases seek both damages and a declaration of enforceable rights over training and outputs.
Others are cutting deals. Reported licenses range from $1-$5m annually with several publishers to a multi-year agreement reportedly worth over $250m for News Corp. Not everyone is onboard; some leaders call for collective action and slower dealmaking to avoid underpricing rights.
The legal questions you'll face
- Training liability: Does ingesting public web content for model training infringe copyright, or fit fair use?
- Output liability: When AI Overviews summarize or quote, do they substitute the work and reduce market value?
- DMCA 1202: Are credit/metadata signals removed or obscured during ingestion or output?
- Contract and implied license: Do robots.txt, site terms, or past practices grant or limit rights for AI use?
- Competition concerns: Does tying search visibility to AI reuse create undue leverage over publishers?
Contracting playbook (if you license)
- Scope precisely: training vs. retrieval-augmented use vs. summaries in search; text vs. images vs. audio.
- Attribution: placement, link requirements, and measurement of referral clicks.
- Usage transparency: model, feature, and geography disclosures; change-control approvals for new uses.
- Compensation: floors, volume tiers, revenue share for queries where your content drives answers.
- Data governance: retention limits, re-training restrictions, purge rights, and audit rights.
- Output controls: guardrails against verbatim reproduction, caching, or long-form substitution.
- Indemnities: IP infringement, defamation, and generative errors; caps aligned to actual risk.
- Term and exit: reversion of rights, wind-down of models containing your data, and monitoring post-termination.
- Most-favored terms: parity with similarly situated publishers as the market resets.
Litigation posture (if you sue)
- Build the substitution story: show traffic and revenue displacement tied to AO exposure.
- Document copying: crawler logs, cache evidence, and model outputs that mirror protected expression.
- Market harm: affiliate declines, ad RPM changes, and reduced subscription funnel volume.
- Remedies: damages, declaratory relief on training and output rights, and injunctive guardrails.
Risk controls you can deploy now
- Robots and headers: restrict known AI crawlers (e.g., GPTBot) where appropriate; understand there's no clean opt-out of AO without collateral SEO impact.
- Snippet controls: use meta and on-page directives (e.g., nosnippet or data-nosnippet on sensitive sections) knowing the trade-offs.
- Attribution beacons: structured data and canonical tags to strengthen source recognition in summaries.
- Terms of service: explicit bans on training, caching, and automated extraction; require attribution and linking.
- Watermarking and test pages: seed markers to detect uncredited reuse in summaries or outputs.
- API licensing: funnel high-value use through paid APIs with rate limits and audit provisions.
International considerations
Challenges are emerging across India, Canada, Japan, and Brazil. Expect differences in exceptions, moral rights, and collective licensing models.
If you operate globally, align contract language and enforcement with local statutes, data residency rules, and consumer protection regimes. One agreement rarely fits all regions.
What to watch through 2025
- Early rulings on training and output liability that shape settlement ranges.
- Regulatory guidance on AI ingestion and attribution standards.
- Deal momentum and whether MFNs start to anchor pricing.
- Product changes from Google that alter how AO cites sources or routes clicks.
Practical next steps
- Run a quarterly AO impact review: queries, CTR shifts, and revenue deltas.
- Prepare dual tracks: a litigation file and a licensing term sheet, so you can move fast either way.
- Coordinate with editorial, SEO, and revenue teams on snippet controls and high-stakes pages.
- Brief the board with scenario ranges: no-deal, license, and litigation outcomes over 12-24 months.
Bottom line
AI Overviews compress demand at the source. The business model for journalism will follow whatever precedent and contracts emerge in the next 12-24 months.
Move early, measure hard, and keep both options open-courtroom and contract. Your leverage depends on the evidence you collect now.
Resources
- U.S. Copyright Office: Artificial Intelligence
- Google Search documentation: AI Overviews
- Complete AI Training: Courses by Job
Your membership also unlocks:
