​

What are people really using ChatGPT for?

If you thought ChatGPT was mostly for work - emails, code, reports - think again. The largest study of its kind, from OpenAI and Harvard economists, analysed 1.5 million conversations and found something very different: nearly three-quarters of all usage is now non-work.

This shift has big implications. People are turning to AI not just as a productivity tool, but as a life advisor.

The data shows three categories dominate: Practical Guidance, Seeking Information, and Writing. Together, they account for almost half of all use. The fastest growth is in Seeking Information, which has doubled in a year - from 14% to 24% of all messages. Think questions like “What are the best running shoes for flat feet?” That’s search-like behaviour, shifting to AI. Customers aren’t just Googling anymore - they’re also asking ChatGPT.

In traditional search, intent has long been divided into navigational, informational, and transactional. OpenAI’s study maps usage into 26 categories - a new intent framework for AI.

Who’s asking is just as revealing. The gender gap? Gone - women now slightly outnumber men. Nearly half of usage is under-26s. And adoption is growing four times faster in lower- and middle-income countries like India, Nigeria, and Indonesia than in wealthy ones. This is mainstream, global, and young.

Work usage is still there. Writing makes up 40% of work-related tasks - but two-thirds of that is editing text, not generating it. Coding? Just 4.2%. Companionship, therapy, role-play? Less than 3% combined.

The real story is advice, information, and practical guidance.
​
Almost half of all messages fall into the “Asking” category - people wanting judgment, recommendations, guidance. Those conversations also get the highest satisfaction scores. People like ChatGPT best when it’s helping them decide, not just do.

And while its usage is still smaller than Google Search, the scale is enormous: 700 million people use ChatGPT every week. That’s about 10% of the world’s adults - more than Netflix or LinkedIn.

Read the summary from OpenAI and the full report.
​
​

​
Meanwhile, elsewhere this week in search and AI...
​​
​
​Google launches Agent Payments Protocol (AP2)​
​
Think of AP2 as the TCP/IP of agentic commerce: an open protocol that allows AI agents to securely buy, sell, and transact on our behalf.
​
Backed by 60+ companies - including Mastercard, PayPal, Coinbase, and Amex - AP2 introduces “mandates”: cryptographically signed proofs of user intent. In practice, that means agent-led purchases can be traceable and auditable across credit cards, bank transfers, and even stablecoins.
​
It’s early days, but this looks like the foundation for how agents will handle money at internet scale.
​
Agent-first commerce is coming. With AP2, buying decisions may never reach your website - they’ll be made inside AI agents. That means optimising for agent discovery, trust, and eligibility, not just search, social, and paid media.
​
Read more on AP2 and agentic commerce:

​
👉 Google Cloud Blog – Powering AI commerce with the new Agent Payments Protocol (AP2)​

👉 [Video] Google + Coinbase video demo – Agent-to-Agent retail with AP2 + x402 payments​

👉 ZDNet - Google’s new open protocol secures AI agent transactions – and 60 companies already support it​

👉 Coinbase Blog - Google Agentic Payments Protocol + x402: Agents can now actually pay each other​
​
​
​Schema in Google’s AI Mode​
​
Salt Agency analysed over 107,000 URLs cited in Google’s AI Mode - the new AI option in Search (a bit like ChatGPT built into Google) - and the results might surprise you.
​

Schema, it turns out, is hygiene, not a differentiator. The most common types showing up are the basics: Organization, WebPage or Article, Breadcrumbs, Author. But those niche types - FAQ, HowTo, Product, Review - don’t correlate with higher visibility in AI Mode. In other words: marking up your Product pages won’t magically make you a cited source.
​

Schema is still worth doing. It helps with traditional SEO - eligibility for rich results and giving Google a better understanding of your site. But when it comes to AI Mode, it won’t win you a place. What matters is the authority and clarity of your content - that’s what gets pulled into AI answers.
​
👉 Salt Agency Blog - Does schema help you surface more in Google’s AI Mode?​
​
​
​Were we wrong about “The Great Decoupling”?​
​
For the past year, SEOs have been pointing to what they call “The Great Decoupling” - impressions in Google Search Console skyrocketing while clicks flatlined. The assumed culprit? AI Overviews stealing traffic.

But new analysis by Brodie Clark points to another explanation: bots - specifically, scraping from SEO tools.

Here’s what happened: Google quietly disabled the &num=100 parameter - the trick tools used to pull 100 search results per query instead of 10. Almost immediately, impressions in GSC dropped sharply, in some cases by hundreds of thousands a day. Which implies that many of those “impressions without clicks” weren’t human at all - they were bots. Rank trackers loading page after page, inflating impression counts.

If that’s true, then AI Overviews’ impact - while real - may have been overstated. And it could also reshape rank tracking for the tools we all rely on, like Semrush and Ahrefs.
​
This resets the conversation on where traffic is really going, and which metrics can still be trusted.
​
👉 Brodie Clark Consulting - Were we wrong about “The Great Decoupling” after all? Analyzing the impact of &num=100​
​
​

​
Google Search glitch floods results with ads
​
​Between 6am–11am BST (1am–6am ET), users worldwide saw SERPs flooded with ads and barely any organic results. Don’t panic - it wasn’t an experiment or permanent change. Google says it was just a serving bug.
​
👉 Google Search bug leads to search results overwhelmed with ads​
​
​​
​
Google hiring an Anti-Scraping Analyst
​
​Google wants an Anti-Scraping Analyst to spot SERP scraping patterns, build ML countermeasures, and shut down the bots inflating Search Console data. Translation: GSC numbers might finally get cleaner.
​
👉 Google is hiring an anti-scraping engineering analyst​
​