ChatGPT Is Starting to Look a Lot Like Google, And That’s No Accident
The line between AI assistants and search engines keeps getting blurrier. This week, OpenAI moved ChatGPT closer to Google’s territory, Anthropic quietly shipped new tools for developers tracking how their AI actually gets used, and a popular AI framework patched a subtle bug that was recording things it shouldn’t. Taken together, these updates sketch out where AI is headed: deeper into everyday commerce, more measurable, and slowly more accountable.
OpenAI Is Letting Businesses Set Up Shop Inside ChatGPT
According to TechCrunch, OpenAI is reportedly rolling out business profiles for ChatGPT, dedicated pages where companies can list their products, services, hours, and contact details directly inside the app. If you’ve ever asked Google for “best pizza near me” and gotten a panel of local restaurants before any search results, you already understand the idea. OpenAI is building something similar, except instead of showing a results panel, ChatGPT can now surface a business’s own information when someone asks a relevant question in conversation.
This matters because it changes what ChatGPT actually is. Right now, when you ask ChatGPT about a local plumber or a software product, it pulls from whatever it learned during training, which may be outdated or incomplete. Business profiles let companies feed ChatGPT current, accurate information about themselves directly. That closes a gap Google has never had, since businesses have been able to manage their Google listings for years.
For everyday users, this could make ChatGPT genuinely useful for local and commercial searches that it currently handles poorly. For small business owners, it raises a new question worth taking seriously: if people are searching AI instead of Google, do you have a presence there?
Why this matters: ChatGPT reportedly handles over a billion queries a week. If businesses need profiles there the way they need Google listings today, an entirely new slice of digital marketing is about to open up.
“OpenAI is adding business profiles to ChatGPT so companies can provide information directly to the app.”
Anthropic’s Developer Tools Now Track How Claude Gets Used
Anthropic released version 2.8.0 of its Python SDK, a software toolkit that lets developers connect their apps to Claude, Anthropic’s AI model. The headline addition is analytics: developers can now track how Claude is being used inside their products, which conversations triggered which responses, and how often Claude’s answers get accepted or rejected by end users.
Think of it like Google Analytics, but for AI. Before this update, a company that built a customer service tool powered by Claude had limited visibility into whether Claude was actually helping their customers or quietly frustrating them. The new analytics features give developers a clearer picture, so they can tune their products based on real behavior rather than guesswork. The update also adds controls for Slack integration, specifically, which channels Claude is allowed to respond in, alongside fixes for a bug affecting Amazon Bedrock, Amazon’s cloud service for running AI models.
For people using apps built on Claude, this update is invisible but meaningful. It means the developers behind those apps can now identify and fix problems faster. Better data for developers usually means a better experience for the people actually using the product.
Why this matters: As AI gets embedded in more business tools, the ability to measure what’s working becomes as important as the AI itself. This release, detailed on GitHub, is a step toward treating AI like any other product that needs to be monitored and improved continuously.
LangChain Fixed a Bug That Was Quietly Recording the Wrong Things
LangChain, an open-source framework that helps developers chain together multiple AI tools into a single workflow, released version 1.4.10 this week with a fix that sounds small but carries real implications. The bug caused internal system calls, background instructions that the AI sends to itself during processing, to show up in the message logs that developers review. Those calls were never meant to be visible or tracked.
Imagine a translator who keeps notes in the margins while working. Those margin notes are useful internally but were never meant to end up in the final document. That’s roughly what was happening here. The update also improves how the framework handles AI refusals, when a model declines to answer something, by stopping unnecessary tracking data from being attached to those refusals.
For developers building complex AI pipelines, cleaner logs mean easier debugging and less noise when something goes wrong. For everyday users, it’s a reminder that the AI tools in their lives are built in layers, and each layer needs its own maintenance. The full release notes are available on GitHub.
Why this matters: Logging and transparency inside AI systems are becoming a compliance issue, not just a technical one. Fixing what gets recorded, and what doesn’t, is exactly the kind of unglamorous work that keeps AI products trustworthy.
Also Happening in AI
ComfyUI, a popular tool for generating AI images through a visual interface, dropped version 0.45.0 with frontend updates and better error handling. LiteLLM, a library that acts as a universal adapter letting developers swap between different AI models without rewriting their code, released v1.106.8. On the regulation front, New York’s AI safety law officially took effect, requiring companies like OpenAI and Anthropic to publish details about their models’ capabilities and risks, a development TechCrunch calls the most significant state-level AI oversight yet. Adobe is experimenting with tools that let you walk around inside photographs and have conversations with characters in your images, according to The Verge. Meanwhile, eBay appears to have blocked third-party AI agents, automated programs that shop or bid on a user’s behalf, from its platform, The Verge reports, raising questions about how marketplaces will handle AI-powered buyers going forward.
What to Watch
The OpenAI business profiles story and eBay’s block on AI agents are part of the same underlying tension: who controls commerce when AI is in the middle of it. Watch for other major platforms to either follow eBay’s lead and restrict AI agents, or move in OpenAI’s direction and build dedicated infrastructure for them. New York’s new AI safety law is also worth tracking closely, if enforcement has teeth, other states will pay attention.