OpenAI made two moves this week that put local business squarely in its sights. On Tuesday it launched the ChatGPT for small business program, a free training and tooling push aimed at owners, not developers. Then on Thursday, Yelp announced it is licensing its reviews, photos, and business data to OpenAI so ChatGPT can recommend local businesses with fresher, more specific answers. Read those two stories together. The biggest AI company on the planet just named Main Street its next growth market, and it wants both your workflows and your reputation data.
OpenAI wants your workflows and your reviews
The program is concrete, not a press release with a landing page. It includes free product webinars, in-person small business AI academies across the US, starter guides, and partner integrations built with Shopify and Intuit, all running on GPT-5.6. OpenAI also disclosed that ChatGPT Work and Codex have passed 10 million business users. Its pilot data claims 78% of Small Business AI Jam attendees built a working automation in a single day, and 42% now save more than five hours a week.
Hold the vendor numbers loosely. A one-day workshop stat from the company selling the subscription is marketing, not research. But the competitive picture is real. Anthropic pushed Claude Cowork to mobile and web this month, OpenAI answered with ChatGPT Work on GPT-5.6, and both are now selling the same promise to non-technical owners: an assistant that finishes multi-step office work, not just drafts. When two rivals fight over your segment with free training, the tools get cheaper and better. Let them fight. Take the free classes, buy nothing on day one.
The Yelp deal is the bigger story for anyone with a storefront or a service area. Yelp’s reviews, images, and business details now feed ChatGPT’s local recommendations directly. Your Yelp profile stopped being a directory listing on Thursday. It became source material for the answer engine that a growing share of your customers ask first. A thin profile, stale photos, or three unanswered one-star reviews now shape an AI’s summary of your business, delivered with total confidence to someone who will never click through to check.
AU’s take: treat this like the Google Maps land grab of 2010, because that is what it is. This weekend, claim the Yelp profile if you never did, fix hours and service lists, load current photos, and write owner responses to your last ten reviews. The model reads all of it. And before your team shows up to a free OpenAI academy, put rules under them first. Monday’s post covered the gap: 77% of small businesses run AI daily and only 23% have any written policy.
Local search is now two contests, and you may be winning only one
Search Engine Journal published numbers this week that should reframe how you read your rankings. AI Overviews now appear on 68% of local searches, while the classic local pack shows on just 39% of the same queries. Plenty of businesses rank fine in the pack and never appear in the AI answer above it. Those are different games with different scoring, and the AI game now runs on more real estate.
Google did not make the week easier. Since July 9, Business Profile dashboards have told some owners they have no reviews at all while their live listings still show every one. Google confirmed it is a display bug and says a fix is coming. At the same time, review enforcement keeps tightening: profiles now get asked directly whether they reward customers for reviews, and non-compliant listings are picking up public warning banners.
Getting cited in an AI Overview is not a mystery, either. The same SEJ piece lays out the pattern behind the businesses that show up: a complete Business Profile, service pages that answer one question each in plain language, and structured data that names the service and the city. That is a weekend of work for most local sites, not a retainer. If your plumber competitor outranks you in the AI answer while you hold the number two pack spot, this is why.
The practical read: do not judge your review program by a broken dashboard, and do not let anyone on your team trade discounts for stars. If you want to know what the AI shift is doing to your actual traffic, Wednesday’s post walked through the new GA4 labels that tag AI assistant referrals as their own channel. Measure it before you react to it.
Real estate ran out of denial room
The July industry dispatches put hard numbers on something agents have felt all year. 67% of buyers now start their property search in a generative AI tool, up from 17% eighteen months ago. Meanwhile 91% of agents never appear in AI answers at all. Adoption is not the problem: 97% of brokerage leaders say their agents use AI, and 82% write listings with it. Visibility is the problem, and it is concentrated in a small group who did the unglamorous schema and profile work.
Zillow’s rentals integration inside Gemini, which went live earlier this month, points at the structural version of the same shift. Portals are becoming databases that AIs query, and the agent relationship starts one layer up from the portal. Tuesday’s post argued the counter position: buyers who start with AI still pay humans to verify, and 44% will pay a premium for an agent who checks the AI’s work. The verification role is real. You just have to be findable by the machine first.
Field service vendors started selling by trade, not by seat
Housecall Pro launched its first trade-specific packages for HVAC, plumbing, and electrical shops, its first vertical offerings after years of one-size-fits-all plans. That is a signal worth reading. The field service platforms spent two years bolting AI onto generic workflows. Now they are packaging it by trade, with the pricing ladder to match: Housecall Pro runs $79 to $329 a month with its CSR AI included, while ServiceTitan sits at $245 to $500 per tech per month at the enterprise end.
If you run trucks, the takeaway is that the mid-market just got more competitive on exactly the features that used to require enterprise money: AI answering, estimate follow-up, and dispatch help. Before you renew anything this quarter, price the trade-specific bundles against what you actually use. And check the intake layer for the leak Thursday’s post covered: Spanish-language calls are hitting 30% in Sun Belt markets and most voice AI setups book none of them.
One question for the weekend
Every story this week is the same story wearing different clothes. OpenAI licensing Yelp, AI Overviews outrunning the local pack, buyers starting home searches in a chatbot, vendors packaging AI by trade. The first answer a customer hears about your business is increasingly written by a machine reading whatever public data you left lying around.
So here is the question to sit with over the weekend: if a stranger asked ChatGPT about your business right now, what would it say, and which part of that answer did you actually choose? Open the app and ask. The gap between what it says and what you would say is your work list for August. We will be here Monday with the next play.
