Alphabet Loses $460B in a Day — and Hands PR a New Scorecard

Alphabet shed roughly $460 billion in market value by midday after its third-quarter report — a one-day wipeout that came close to Nvidia's record loss of about $593 billion, set when DeepSeek rattled the AI trade in January 2025.

That's the headline number. But the numbers that will matter to communications practitioners are smaller and more specific, and they describe a shift already underway in how buyers find brands.

What spooked investors?

The sell-off is about spending, not revenue collapse. Capital expenditure hit $51.4 billion, free cash flow went negative by $9.1 billion, and earnings missed expectations.

CEO Sundar Pichai used the earnings call to explain where the money goes: Google is building one unified experience out of AI Overviews, AI Mode and traditional search. Queries in AI Mode run three times longer than traditional searches, and one in five uses voice, images or documents. The Gemini app has passed a billion users, and people bring it multi-step tasks, not just single questions.

The takeaway: Google is not bolting AI onto search — it is rebuilding search as a conversation.

What happened to search revenue?

Search grew 12% to $63.3 billion. That sounds healthy until you compare it to last quarter, when growth ran 17%. Management said lower monetization in AI Mode weighed on the result.

In plain English, the company that sold the top of the results page is rebuilding it as an answer — and an answer does not sell like a page of links. A results page had ten slots and sold the first few. An answer names a handful of brands, and the buyer trusts the answer more than the ad beside it.

Google will sell ads inside answers, and it has already said so. But the credibility layer — the part that decides who gets named — draws on sources the tool trusts. That is earned media, whether or not the budget line says so.

Why do longer questions change the sourcing game?

A buyer who asks a three-sentence question gets a three-sentence answer, built from whatever the tool finds that addresses the whole question. Thin pages and recycled press releases don't make that cut.

What does get read, according to 5W founder and chairman Ronn Torossian, whose analysis of the earnings report framed this story:

• Independent coverage in publications the tools trust
• A client's own pages that state facts plainly
• Video and community conversations — including Reddit — that the tools ingest

Torossian's framing: PR has always been a mix of journalism, psychology and lawyering, and each part counts twice now. Journalism gets you into the publications the tools read. Psychology persuades the editor and then the reader. Lawyering matters because a machine repeats what it finds — a loose claim can be quoted back in answers for years.

The takeaway: whatever an AI assistant says about your client tomorrow is being sourced today.

How should agencies change the scorecard this quarter?

Torossian's prescription is blunt: placements and impressions no longer tell a client whether ChatGPT, Gemini, Perplexity or Copilot names the brand.

His suggested test:

• Pick 20 questions your customers actually ask
• Run them monthly through the major answer engines
• Count brand mentions, note which publications the answers cite, and check whether the tools' statements are accurate

An agency that resists that test should be asked why, he argues. The firms that struggle will be the ones that split the work into pieces nobody owns — the press-only shop, the dashboard reseller, the impressions reporter. At 5W, he notes, media, content, creative, video, a dedicated Reddit team and proprietary technology sit in one place, with a human editor reviewing every piece before it goes live — and he would ask any agency, his included, who owns the work end to end.

The trade-press caveat applies here: that last paragraph is a vendor making a vendor's case. The audit method, though, costs nothing to run and answers a question clients are already asking.

What's the bottom line for practitioners?

Alphabet just spent roughly half a trillion dollars of investor patience to signal where buyers are heading — toward longer, multi-modal queries answered by machines that name a shortlist of brands.

Count names, not reach.