What Happened This Week
Google spent this period pushing AI search deeper into new territory – a contested French launch, tighter ChatGPT Ads mechanics from OpenAI, a market that isn’t moving on the surface – while the relationship between platforms and the publishers who supply their answers kept fraying in public. USA Today’s CEO said blocking Google is a live option. People Inc’s CEO called it “100% on the table.” Reddit is reconsidering the licensing deal it signed with Google in 2024. France’s publisher groups say Google launched AI Overviews there without an agreement and are asking their government to intervene. None of this slowed anything down – Google shipped the France rollout on schedule, OpenAI kept building out its ad platform, and the EU’s regulatory actions this period didn’t actually touch AI Overviews despite some reporting suggesting otherwise. The structural story isn’t confrontation stopping deployment. It’s deployment continuing while the economics underneath it get more openly contested.
The Findings
Did Google’s France launch resolve its publisher dispute, or just relocate it?
Google launched AI Overviews and AI Mode in France on July 22, 2026, across desktop, mobile, and the Google app – announced by Elizabeth Reid, VP of Search, via a LinkedIn post, two months ahead of an originally communicated September 23 deadline for French publishers. The launch sits on top of real regulatory history: France’s Autorité de la concurrence fined Google €250 million in March 2024 (Decision 24-D-03) for failing to honor 2022 commitments to negotiate neighboring-rights payments with press publishers, and Google’s existing neighboring-rights deal – covering roughly 300 press titles – was renewed in January 2025.
What was not confirmed: initial research characterized this as a settled arrangement, including a specific claim that Google was paying “450 contracted press outlets” under a new settlement framework tied to the launch. That does not hold up. Contemporaneous French reporting (Journal du Net, Graphiline) describes Google deploying without a new publisher agreement specific to this launch – the French press alliance called it a “fait accompli.” The magazine publishers’ union SEPM is actively disputing the rollout, says it currently lacks the data to assess its impact, disputes the opt-out mechanism, and has asked for government intervention. Treat any claim of a formal 2026 settlement as unconfirmed.
Practitioner implication: if you manage search presence for a French or EU publisher, don’t assume the opt-out toggle (Search Console > Settings > “Search generative AI”) is protecting your commercial position just because it exists – the people it’s meant to protect are actively disputing whether it does. Watch Autorité de la concurrence and SEPM statements directly rather than Google’s own framing of the rollout.
Is the EU actually going after AI Overviews, or just Search rankings and the Play Store?
On July 23, 2026, the European Commission issued its first Digital Markets Act non-compliance fine against Google – €890 million total (€460M for Search self-preferencing, €430M for Play Store anti-steering). Separately, on July 24, Google signed the EU AI Act’s voluntary Code of Practice on Transparency of AI-Generated Content, committing to disclosure and SynthID watermarking for AI-generated content generally.
What was not confirmed: the July 23 fine does not name or target AI Overviews or AI Mode – reporting on the fine is explicit that “the investigation predated their large-scale deployment.” A separate Commission investigation, opened December 9, 2025 (outside this period and still open), does explicitly examine whether Google’s use of publisher and YouTube content to power AI Overviews and AI Mode breaches competition law – that’s a distinct, unresolved proceeding, not something this fine settles. The Transparency Code, similarly, is a general AI-content labeling commitment, not a search-specific mandate.
Practitioner implication: don’t conflate this fine with AI Overviews enforcement – it’s about ranking self-preference and app-store steering, unrelated to your AI Overviews visibility. The actual AI Overviews antitrust question is still pending separately at the Commission; that’s the one to watch for real impact on how Google can use publisher content.
Are US publishers actually about to block Google, or is this posturing?
A Wall Street Journal report (July 21, 2026, corrected the following day) put a spotlight on publisher frustration with declining referral traffic in the AI Overviews era. USA Today Co. CEO Mike Reed said the company already blocks crawlers without licensing agreements and said “it’s time to take a stand” if declines continue. People Inc CEO Neil Vogel said blocking Google entirely is “100% on the table.” Reddit is re-evaluating its roughly $60 million/year content-licensing deal with Google ahead of renewal (CNBC, Neowin, Winbuzzer, July 22-23). Politico employees have discussed requiring registration or login to limit Google’s bot access. Reuters is separately, and more tentatively, evaluating whether to block Google’s bot from consumer-facing news – a Reuters Digital executive said the company lacks enough data to decide quickly. The Economist’s VP of Generative AI, Josh Muncke, said the company has examined opting out of AI features specifically, though a full Google exit “isn’t currently part of that discussion.”
What was not confirmed: the WSJ’s original traffic-decline figures were corrected the day after publication – Business Insider’s cited decline, for instance, was revised down from over 85% to 43%. The actual range across named outlets runs roughly -20% to -43% year-over-year, not a uniform “40%+ across major publishers” – and it isn’t universal: the Guardian and BBC reportedly gained traffic (+16% and +15% respectively) in the same period. Treat any single “publishers are losing 40%+” figure as an overgeneralization of a much wider, mixed range.
Practitioner implication: this is escalated talk, not confirmed action – no major publisher had actually pulled Google access as of this writing. But CEOs saying this on the record, not anonymously, is a real signal. The concrete decision point to watch is Reddit’s deal renewal, not the more rhetorical USA Today/People Inc statements.
What did OpenAI just change about ChatGPT Ads, and why does it matter that there was no public announcement?
On July 24, 2026, OpenAI communicated eight updates to ChatGPT Ads directly to registered advertisers via private correspondence, not a public post. Confirmed changes: a Conversions objective for conversion-optimized bidding; Automatic Advanced Matching using hashed identity data; a shift from fixed daily budget caps to seven-day rolling/average budget pacing; geographic exclusion targeting; a bulk asynchronous API for campaign management; and refreshed shopping cards with dynamic pricing and star ratings.
What was not confirmed: no public announcement accompanies these changes as of this writing – all confirmation comes from advertiser correspondence surfaced by trade press (Search Engine Land, MarTech, ppc.land, Search Engine Roundtable), not an OpenAI-published source.
Practitioner implication: if you or your clients run ChatGPT Ads, this changes budget pacing and bidding mechanics without a public changelog. Check accounts directly rather than waiting for an OpenAI blog post – these changes are shipping via private advertiser communication, not announcement.
Where does search market share actually stand heading into August?
StatCounter’s July 2026 data puts Google at 91.31% global search market share and Bing at 4.47%.
What was not confirmed: this is a snapshot, not a trend line – we found no confirmed event in this period tied to the month-over-month movement.
Practitioner implication: raw market share isn’t where the competitive action is this period. The real movement is beneath these numbers, in commerce integration and citation behavior across platforms – see the next finding.
How often does an AI Mode ad actually earn you a citation?
SE Ranking analyzed over 50,000 commercial keywords in Google’s AI Mode, published July 14, 2026 (data collected through June 30, 2026 – just before this period technically opens, included here as the closest and best-sourced current data on AI Mode commerce mechanics). Ads appeared in roughly 29% of commercial AI Mode queries overall. Ad presence scaled sharply with keyword value: under 25% for low-CPC keywords, over 50% for keywords at $10+ CPC. Category variance was extreme – pet care queries showed ads around 72% of the time, healthcare under 3%, likely reflecting YMYL caution. Buying an ad placement earned the advertiser an organic source citation in the AI Mode answer only about 11.5% of the time.
What was not confirmed: the study’s stated figure for how often an advertiser also held an organic ranking on the same page – “2.3%” – doesn’t match SE Ranking’s own published numbers, which imply a considerably higher share (roughly 15%) held some organic ranking. Use the ad-presence and citation figures with confidence; treat the organic-ranking-overlap figure as unconfirmed.
Practitioner implication: paying for an AI Mode ad placement is not a shortcut to earning a source citation – the two are largely decoupled. If a citation strategy depends on ad spend, this data says that’s not how it works; the organic/GEO fundamentals still determine citation, regardless of ad budget.
What This Means For You
Nothing this period suggests deployment is slowing to accommodate the publisher relationship problem – it suggests the opposite. Google shipped France on schedule despite an unresolved publisher dispute. OpenAI kept building out ad infrastructure while its research and ads teams operate largely outside public view. The EU’s most visible enforcement action this period didn’t even touch the AI Overviews question that’s actually contested at the Commission. What’s changed is that the publisher side of this relationship is now negotiating in public rather than privately – on-the-record CEO statements, not anonymous sourcing – and regulators are drawing sharper, if still fragmented and slow-moving, lines around who’s liable when an AI summary gets something wrong. If you’re advising a publisher or a brand right now, the operative question isn’t whether platforms will pull back. It’s how much leverage the public pressure campaign actually buys before the next licensing deal or opt-out toggle gets tested for real.
What to Watch Next
- Reddit’s Google licensing deal renewal – In testing/evaluation. Reddit is actively weighing tighter data access terms or non-renewal; no decision confirmed yet (CNBC, Winbuzzer, July 2026).
- Politico and Reuters crawler-access decisions – Announced/discussed, not yet decided. Both are evaluating limits on Google’s bot access; Reuters says it lacks sufficient data to move quickly.
- The EU Commission’s separate AI Overviews antitrust investigation – In progress. Opened December 9, 2025, still open as of this briefing; this is the proceeding that actually targets AI Overviews and AI Mode content use, distinct from July’s DMA fine.
- French publisher groups’ response to the AI Overviews launch – In progress. SEPM and the Alliance de la presse d’information générale are actively contesting the rollout and have requested government intervention; watch for a formal regulatory response.
- Google’s appeal of the Munich Regional Court liability ruling – In progress. The court ruled May 28, 2026 (reported prominently in early June) that AI Overviews generate independent, substantive statements and Google can be liable for false ones regardless of source links; Google is appealing, so this is not yet settled precedent, but it’s the clearest legal signal so far on platform liability for AI-generated inaccuracies.
- Ahrefs’ updated organic-click-reduction figure – Confirmed published (pre-period). Ahrefs has superseded its widely-cited “34.5% click reduction” figure (April 2025) with an updated 58% figure for a more recent measurement window – if you’re citing Ahrefs on AI Overviews’ click impact, use the current number, not the older one still circulating.
Did the EU’s July 2026 DMA fine target Google’s AI Overviews?
No. The €890 million fine the European Commission issued on July 23, 2026 covers Search self-preferencing and Play Store anti-steering – it does not name or target AI Overviews or AI Mode. A separate Commission investigation, opened December 9, 2025 and still open, is the one actually examining whether Google’s use of publisher content to power AI Overviews and AI Mode breaches competition law.
Is Reddit ending its data-licensing deal with Google?
Not yet. Reddit is actively re-evaluating its roughly $60 million/year content-licensing deal with Google ahead of renewal, following declines in search-referred traffic, but no decision to end or renew it had been confirmed as of this writing.
Does buying an ad in Google AI Mode get you cited as a source?
Rarely. A July 2026 SE Ranking study of over 50,000 commercial AI Mode keywords found that advertisers earned an organic source citation in the AI-generated answer only about 11.5% of the time – ad spend and citation are largely decoupled.
Did French publishers agree to Google’s AI Overviews launch in July 2026?
No. Google launched AI Overviews and AI Mode in France on July 22, 2026 without a new publisher agreement specific to the rollout. French publisher groups, including the magazine publishers’ union SEPM, are actively disputing the launch and have requested government intervention.
Can Google be held legally liable for false information in AI Overviews?
Yes, according to a Munich Regional Court ruling that Google is currently appealing. The court held that AI Overviews generate independent, substantive statements – not neutral indexing – making Google potentially liable for false assertions regardless of whether a source link is included. The ruling is not yet final precedent.
