Judge Tossed the Google AI Overviews Suits. Now What?

By Lazar Milicevic · Published October 6, 2026 · 8 min read

If a meaningful share of your leads come from Google, the court just told you nobody is going to fix that dependency for you. Two companies with real legal budgets, Chegg and Penske Media, tried to hold Google accountable for AI Overviews and lost. Here's exactly what the ruling said, what it doesn't settle, and a 90-minute audit you can run on your own site this week.

What the judge actually decided (and what he didn't)

On September 30, 2026, Judge Amit P. Mehta of the U.S. District Court for the District of Columbia signed a 41-page memorandum opinion dismissing both the Chegg and Penske Media antitrust suits over AI Overviews. Mixed News reports that he also declined to hear the California unjust-enrichment counts, so those were never decided on the merits.

The core reasoning is worth reading closely, because it's the part that applies to you. The publishers pleaded an expectation of search traffic, not an agreement. Per Press Gazette, Mehta wrote that an expectation is not an agreement, and that Google sending traffic is simply how a general search engine works. Your rankings were never a contract.

Two more findings matter:

• Search and AI Overviews are one product. Penske's tying claim failed because the court treated Google Search and AI Overviews as a single integrated search experience, not two separate products (Neoteo). You can't assume AI summaries will be treated as something separate from ordinary search.
• Lost clicks aren't an antitrust injury. The court found that lost referral traffic and ad revenue are not direct antitrust injuries caused by Google, and that the publishers lacked antitrust standing in general search because they operate downstream of Google (Technology Org).

What the ruling does not do: it doesn't decide whether Google's use of publisher content violates copyright law. According to Media Copilot, it closes only these particular antitrust arguments in district court. The order is final and appealable, and as of early October 2026 I haven't seen an appeal announced. I'm not going to speculate on what happens next. If you want the legal detail, read the opinion itself.

One framing note. The lost-traffic figures floating around in coverage of these cases are the plaintiffs' allegations, not court findings. Don't treat them as measured facts about your own situation either. Measure your own.

Why the ruling changes your risk, not just publishers' risk

The practical takeaway for a small business: stop planning around "Google will have to fix this." That sentence is a hope, not a plan. Treat AI summaries on the results page as a permanent feature of how people find things and design around it.

Google's own numbers explain why. In its announcement for website owners, Google says AI Overviews has over 2.5 billion monthly active users and AI Mode has passed one billion monthly users. This isn't an experiment sitting at the edge of search anymore.

The risk isn't spread evenly across your site. Think about which of your pages exist to answer a question:

• How much does X cost?
• What's the difference between X and Y?
• How do I do Z?

Those are exactly the pages a summary can absorb. The reader gets the answer on Google and never visits. Pages that sell a service, show real work, or capture a lead are much harder to replace, because a summary can't book the call or deliver the result.

And underneath that is the real exposure: dependency. If one channel sends most of your leads and you don't own the relationship, you have a single point of failure. This pattern is common among small businesses that rely on a single channel. Picture an invoicing business that gets all its leads from search, with no email list, no follow-up, and no way to reach anyone who visited and didn't buy. When clicks drop, they have nothing underneath. (That's a composite of a common pattern, not a named customer.)

The 90-minute traffic dependency audit

This is the working part. Four steps, one spreadsheet, no special tooling.

Step 1: List your top 20 pages by search visits. Pull this from your analytics for the last 90 days. Landing page, organic search sessions, nothing else.

Step 2: Label each page. Two buckets:

• Answer pages: the visitor's question is resolved on the page (pricing explainers, comparisons, how-tos, definitions).
• Move pages: the page moves someone toward a purchase or conversation (service pages, case studies, booking pages, product pages).

Step 3: Score the exposure. For each answer page, note how much of your total organic traffic it drives. A simple rule I use: any answer page carrying more than 5% of your organic visits gets a capture mechanism this week. That 5% is my own working threshold, not an industry benchmark, so adjust it to your volume.

Step 4: Add capture and automatic follow-up to every flagged page. Capture means something specific to the page:

A throwaway spreadsheet is enough to run the audit. Here's a minimal script if you'd rather classify from a CSV export:

import csv

# analytics_export.csv columns: page, organic_sessions
ANSWER_HINTS = ("how-to", "what-is", "vs", "cost", "price", "guide", "faq")

rows = list(csv.DictReader(open("analytics_export.csv")))
total = sum(int(r["organic_sessions"]) for r in rows)

for r in sorted(rows, key=lambda r: -int(r["organic_sessions"]))[:20]:
    share = int(r["organic_sessions"]) / total * 100
    kind = "ANSWER" if any(h in r["page"] for h in ANSWER_HINTS) else "MOVE"
    flag = "<-- add capture" if kind == "ANSWER" and share > 5 else ""
    print(f"{share:5.1f}%  {kind:6}  {r['page']}  {flag}")

URL-slug matching is crude. You'll override a few labels by hand, and that's fine. The point is a ranked, labeled list in front of you, not a perfect classifier.

The follow-up step is where small teams stall

Capturing an email is easy. What happens in the next ten minutes is where most small teams stop, because it means wiring your site, your inbox, and your CRM together, usually by hand.

The target behavior is simple: a new lead lands, gets tagged by the page they came from, receives a useful email within minutes, and shows up in your pipeline. No manual copy-paste. Here's the skeleton of that flow:

trigger: form_submitted
steps:
  - tag_lead:
      source_page: "{{ form.page_url }}"
      topic: "{{ form.lead_magnet }}"
  - send_email:
      template: "deliver_{{ form.lead_magnet }}"
      within: "5m"
  - create_crm_record:
      stage: "new"
      owner: you
  - schedule:
      email: "followup_day_3"
      stop_if: "lead.replied or lead.booked"

Two design rules I hold to when building these:

• Stop conditions are mandatory. A follow-up sequence that keeps emailing someone who already booked a call is worse than no sequence.
• Tag by source page. When Google changes what it shows, you want to know which pages were feeding your pipeline, not just that "organic" dipped.

How long a setup like this takes depends on your tools and how many integrations you need to connect. Once it's running, a traffic drop becomes an inconvenience instead of an emergency, because you're no longer starting from zero every time a visitor leaves.

What Google's controls give you (and what they cost)

You do have some levers, and it's worth knowing exactly what they do, because several are widely misunderstood.

Eligibility is the same as normal search. Per Google's documentation, a page must be indexed and eligible to appear in Search with a snippet to be a supporting link in AI Overviews or AI Mode. There is no separate AI index and no extra requirements.

Snippet directives work, with a catch. Google's robots meta tag documentation says nosnippet applies to AI Overviews and AI Mode and prevents the content from being used as a direct input; max-snippet limits how much may be used. The catch, as PinMeTo notes: the same directives that pull content out of AI answers also remove the snippet from ordinary search results. You're trading visibility for control.

<!-- Removes the snippet everywhere, including normal results -->
<meta name="robots" content="nosnippet">

<!-- Caps how much text Google can use -->
<meta name="robots" content="max-snippet:50">

<!-- Hide one passage only -->
<p>Public text. <span data-nosnippet>Excluded text.</span></p>

Google-Extended is not the AI Overviews switch. PPC Land explains that the Google-Extended robots.txt token governs Gemini training and grounding. It does not keep content out of AI Overviews or AI Mode. Plenty of people block it believing it does. Check Google's official crawler documentation for the current wording before you rely on it.

There's a Search Console opt-out, with caveats. On June 3, 2026, the UK CMA issued a conduct requirement for AI-search opt-outs, and Google began testing a Search Console toggle covering AI Overviews, AI Mode and AI Overviews in Discover (Tech Times). Google says the toggle is not a ranking signal for regular search. It's reported as a property-level setting under Settings, with page-level controls due by March 2027 (Search Engine Journal). Rollout has been uneven, so I can't promise your property has it. Look under Settings.

Search Console's first AI reports also have a gap: they show impressions and dimensions but omit clicks, click-through rate and search-term data. You can see AI exposure but not its click value.

So should you opt out? For most small businesses, my read is no, not blindly. If an AI answer cites you, that's exposure you didn't pay for, and you can't measure what you'd lose. Opting out makes sense only for specific content you'd rather not have summarized, such as paid material, and even then the tools are blunt. Measure first; block second.

Where the follow-up automation fits

Follow-up like the flow above is exactly the kind of repetitive work that practical AI automation can take off a small team's plate. If you'd rather build the flow yourself, the YAML skeleton above is a fair starting point, and you can find more of my work at bizflowai.io.

My take: the lawsuits were the wrong fight for most of us

Even if Chegg and Penske had won, a ruling doesn't give you back a click. What gives you resilience is owning your audience: a list, a pipeline, and a process that works whether the algorithm is kind or not. The ruling just removed the last excuse for waiting.

Run the audit this week. Ninety minutes, one spreadsheet, and you'll know exactly which pages are carrying risk and which ones are already doing the job of moving someone toward a conversation.

Want more like this?

I publish practical AI automation and GenAI engineering content every week.

Planning an AI automation project or need a second opinion on your architecture?

Connect with me on LinkedIn — Lazar Milićević, senior engineer.

Visit bizflowai.io for more on practical AI automation for solopreneurs and small teams.