Google's own AI answers just cost Google about half a trillion dollars. On October 29, Alphabet stock fell close to 13%, its worst day ever, and one of the reasons a senior executive gave investors was blunt: Google's AI Mode shows fewer ads. The next day Reddit lost 29% of its value for the mirror-image reason. Its traffic from Google is drying up.
If you sell online, that pair of headlines is the one to read this weekend. The rest of today's list follows the same money: OpenAI raising less than it wanted, ChatGPT changing how it counts your usage, Gemini renaming the tools you may have built, and Apple finally bringing its AI to China with a local partner.
Google's AI answers are eating Google's own ads
Alphabet reported its third quarter and the market hated it. Free cash flow came in at minus $9.07 billion because of huge AI spending, against analyst hopes of a positive number. The stock dropped roughly 12.6% to 13% in a day, wiping out around $500 billion. That beats its previous worst day, from March 2020.
The detail that matters for you: an Alphabet executive told investors that part of the weakness in search ad revenue came from AI Mode, Google's chat-style search. When people get a full answer inside Google, they click fewer ads and fewer links. Google's answer, as we covered on Friday, is to put ads and shopping suggestions right inside the AI conversation.
Then came Reddit. On Friday its shares fell 29%, the biggest drop in its history, after it warned that Google's AI search was cutting referral traffic and that daily visitors could fall by a low double-digit percentage this quarter.
Why it matters for sellers: "rank on Google and wait for clicks" is getting weaker every month, for everyone, including the biggest site on the internet's front page. Two practical moves:
- Feed the answer boxes. Clear product pages with real prices, sizes and delivery times are what AI search quotes. Vague pages get skipped. We wrote about Google putting a buy button inside Gemini last month; this is the same road.
- Own the conversation. Followers, DMs and WhatsApp chats don't depend on a search engine deciding to send you traffic. Every person who messages you is a lead you don't have to rent. Here's how to turn followers and DMs into real leads.
The AI money is getting nervous
The Google drop wasn't alone. The Financial Times reported on Friday that OpenAI cut the target for its new private round from about $30 billion to about $20 billion, at a valuation of $1.25 trillion instead of $1.4 trillion. The round hasn't closed, so the final number could move again. The Nasdaq 100 has slid around 10% since its October 6 record as investors ask when all this AI spending pays back.
The spending itself isn't stopping. Microsoft trimmed its capital spending forecast for the year to $170 billion from $175 billion, but not by choice: its finance chief said delays from chip suppliers push some new capacity out to the first half of 2027. Alphabet held its own plan at $195 to $205 billion.
Why it matters for sellers: the cheap AI you use today is partly paid for by investors who are now asking harder questions. Two things follow. Prices and usage limits will keep shifting, and capacity stays tight into next year, so busy hours can mean slower tools. Don't panic, and don't sign up for anything on the promise that this month's price is forever. If you're doing the sums on automation, our guide to customer support automation cost walks through the math.
ChatGPT will count your usage in one pot from December
OpenAI's help center now says that starting in December, Plus, Pro and Business subscribers get one shared allowance for ChatGPT and Codex, its coding agent. Usage is counted in tokens, so a heavy model on a long task eats more than a quick question. Instant chat doesn't count against the pot. You'll see what's left, and when it resets, under Settings, then Usage.
OpenAI hasn't published the amounts per plan or an exact start day. It says most people shouldn't hit resets more often than they do now. This comes right after last week's U-turn, when it dropped a planned Pro cut and doubled some allowances instead.
Why it matters for you: if your team uses ChatGPT to write captions, product descriptions or replies, and someone else uses Codex for your website, you'll soon be drawing from the same tank. Agree now who uses the heavy models for what, so nobody finds the account empty on a launch day.
Gemini's October drop: Gems become Skills
Google released its October batch of Gemini features on October 30, eleven in all. The one to act on: custom Gems move to "Skills" from November 17. Skills are saved instructions you call with a slash, like a shortcut. If you built a Gem that writes your product captions in your brand voice, check it after the switch.
Other items worth a look: a Gemini app for Mac with Google's Antigravity coding agent built in (English only for now), file creation and editing inside chat, and more connected apps. The Gemini 4 Argon model, out since October 13, sits next to 3.1 Pro rather than replacing it.
Google Workspace got two small things that are useful in a shop. In Sheets, Fill with Gemini now adds citations to the web sources it used, cell by cell, so you can check where a supplier price or spec came from. In Slides, you can edit a whole presentation from one prompt in the side panel. Handy for catalogs and wholesale decks.
Apple Intelligence reaches China, with Alibaba's Qwen inside
Apple confirmed its first Apple Intelligence features launch in mainland China with iOS 27.3 in December, with a developer beta from November 4. Alibaba's Qwen model powers parts of it, including Siri, Writing Tools and Visual Intelligence. China's internet regulator approved the setup back in July. The new, smarter Siri shown at WWDC is not part of this launch.
Why it matters: it's the clearest sign yet that big AI products arrive in tightly regulated markets through local partners and local models, not as one global copy. For sellers in markets like ours, that's the realistic picture: the AI your customers use will often be a local mix, and the tools that serve you best will be the ones built around how people in your market actually buy.
Quick takes
- Gemini API: Google's Computer Use tool, which lets an AI operate a screen, is now generally available for developers.
- OpenAI developers: the old v1/prompts API and reusable prompt objects shut down on November 30. If a freelancer built something for your shop on them, ask them this week.
- AI detectors misfire: NPR reports writers being wrongly flagged as using AI by imperfect detection tools. If a marketplace or client "detects AI" in your listing, ask what tool they used before you rewrite everything.
- Market mood: the Nasdaq 100 is on track for its first monthly loss since July, mostly on AI spending worries.
The thread running through today
Strip away the stock charts and the lesson is simple. The big platforms are rebuilding themselves around AI answers, and the traffic they used to hand out for free is shrinking. The sellers who do well next year will be the ones who answer their own customers quickly, in the apps those customers already use, and turn each conversation into a sale.
That's the job Vardast does: an AI sales and support assistant that replies to your customers on Instagram, WhatsApp, Telegram and your website, with your real products and prices, around the clock. See you tomorrow.