OpenAI ships always-on agents while Google pays publishers in pennies
OpenAI put autonomous agents on their own cloud computers and cut the price of computer use by half. Google's publisher payment pilot turned out to pay most sites under $1,000, and Google Maps moved review text behind a sign-in.
This Week's Signals
OpenAI launched dots, agents that run continuously on their own cloud computers
Why it matters for your score
A dot can work for days on one task and connect to more than 4,000 apps, which means something reads your site, forms an opinion and moves on without a human ever seeing the page.
GPT-6.1 Sol beats its predecessor on computer use at under half the cost
Why it matters for your score
Agent browsing just got cheaper per task, and cost is what decides how often agents crawl you rather than guess about you.
Google's AI contribution pilot pays about 100 publishers, several under $1,000
Why it matters for your score
Being the source behind an AI answer is not yet a revenue line for anyone but the largest sites, so treat citation as distribution and price your business accordingly.
Google Maps now requires a signed-in account to read or search all reviews
Why it matters for your score
The evidence of your reputation is getting harder for crawlers to reach on Google's surface, which makes reviews you publish on your own domain the more readable copy.
ChatGPT Ads added product review status, bulk feed uploads and 7 new markets
Why it matters for your score
Your catalogue now passes through an approval queue inside ChatGPT, so feed quality decides eligibility before anything decides ranking.
DoorDash launched an ordering agent you reach by text message
Why it matters for your score
The buying surface moved into a message thread with no page to rank on, which is what it looks like when a platform owns the agent and you only supply the inventory.
Emerging Audit Signals
Agent capability is getting cheaper faster than it is getting smarter.
Cost per computer-use task now predicts how much non-human traffic hits your site better than any ranking metric.
Platforms are closing their own data while expecting yours to stay open.
Google now gates review text behind a login, and your product feed still has to be flawless to get in front of anyone.
Getting cited in an AI answer pays in attention, not cash.
Every publisher payment experiment so far rounds to nothing outside the top tier.
The transaction keeps drifting away from the web page.
Text threads, voice calls and in-assistant checkout all route around the site you built for it.
Approval queues are the new gatekeeper.
ChatGPT reviews your products, Google verifies your contact details, and both decide eligibility before visibility is even on the table.
What to do this week
- 1
Load your 5 best commercial pages with JavaScript disabled and read what is left. If price, availability and contact details are missing from the raw HTML, an agent reading you at scale gets nothing. Run /tools/ai-readiness-scan if you want the crawl-access side checked for you.
- 2
Audit your product feed against the attributes AI shopping surfaces require, not the ones that merely pass validation. A rejected product is invisible rather than ranked low, and that rejection now happens inside ChatGPT as well as Merchant Center.
- 3
Republish your 10 strongest reviews on a page you control, with Review schema and the reviewer name and date in the HTML. With Google asking for a login before it shows review text, your own domain is the most reliable place a crawler can read your proof.
Sources
- ↗ Introducing dots, OpenAI
- ↗ Introducing GPT-6.1 Sol, OpenAI
- ↗ Google pays about 100 publishers for AI answers, some under $1,000
- ↗ Google Maps Now Requires Sign In To Read All Reviews & More
- ↗ OpenAI ChatGPT Ads Gains Bulk Product Creation, Product Review Status & More
- ↗ DoorDash launches an AI agent you can text to order food