Google Keeps Fighting AI Web Scrapers After Court Loss. Here's Why It Matters.
Google sued SerpApi for scraping and reselling its search results. A judge handed Google a setback last week, and Google says it's pressing on anyway, with Reddit quietly along for the ride.
Key points
- Google filed a lawsuit against SerpApi in December 2023, accusing it of scraping and reselling Google search results.
- Google used the DMCA, the Digital Millennium Copyright Act, a law normally aimed at DVD pirates and streaming hackers, to argue SerpApi broke through its anti-scraping defences.
- A court ruled against Google last week, but Google confirmed it won't drop the case.
- Reddit joined as a named party with its own interest in blocking unauthorised scraping.
- The outcome could reshape how AI companies collect training data, and who pays for it.
Google took a hit in court last week. A judge ruled against it in a lawsuit against SerpApi, a company that pulls data from Google search results and packages it for businesses to use automatically, at scale, without going through Google directly. Google calls that theft. SerpApi calls it reading public web pages.
Google confirmed it won't walk away.
What exactly did SerpApi do?
SerpApi sold what it marketed as a "Google Search API", software letting paying customers query Google results without Google's involvement or permission. Google argued SerpApi broke through the technical barriers it had built to stop automated harvesting of exactly that kind.
The lawsuit was filed under the DMCA, a 1998 law that makes it illegal to bypass copy-protection systems. Courts have applied it to DVD encryption and pirated streaming. Google argued its anti-scraping technology deserves the same protection.
The content at stake goes beyond blue links. Google pointed to its "knowledge panels", the information boxes appearing in results for well-known people or companies, some of which are built from content licensed from third-party rights holders. SerpApi's scraping, Google said, threatened those licensing relationships.
Reddit, which licenses its own content and faces its own scraping pressures, joined the case as an interested party.
Why does a copyright law designed for DVDs apply here?
That's exactly what the court questioned. Applying the DMCA to a web scraper is unusual, and the judge's scepticism last week shows how untested the theory remains. Google is asking a court to treat anti-bot software the same way it treats DVD encryption. No court has firmly agreed to that yet.
If Google eventually wins, every major website would gain a powerful legal tool to stop AI companies training on their content without permission. We covered the related pressure building on publishers in our 24 July story on what Google's traffic retreat means for the web.
What does this mean for you?
Nothing changes for most people today. But this case sits at the centre of a fight that shapes how AI tools get built. AI models learn by processing vast amounts of text scraped from the web. If courts decide that scraping search results breaks copyright law, building those tools gets meaningfully more expensive.
Higher costs for AI companies tend to flow to users: slower updates, pricier subscriptions, fewer competitors. The Anthropic copyright settlement we reported on 21 July, paying authors $3,000 per book, is an early data point on what that bill might look like.
As Ars Technica first reported, this case is one of several live tests over who owns the data that trains AI.
My read: the DMCA angle is a long shot, and Google knows it. What Google's really doing is making scraping expensive to litigate, which may deter competitors even without a clean legal win. If you publish content professionally, watch the knowledge-panel argument specifically. That's the thread most likely to pull something loose.
