Guides

How to get your business recommended by ChatGPT

When someone asks ChatGPT for the best option in your category, it names a handful of businesses and explains why. Getting into that handful is not luck and cannot be bought. Here are the two systems behind ChatGPT recommendations and the concrete work that influences each one.

Updated August 20, 2026 · 10 min read

To get recommended by ChatGPT, you have to win one or both of two separate systems. The first is conversational recommendation: when someone asks "best physio in Austin" or "which invoicing tool for a small agency," ChatGPT blends what it learned in training with live web retrieval and names a few options. The second is ChatGPT Shopping: a feed-driven product surface that shows shoppable carousels for buying-intent queries. Different inputs, different fixes, and most advice online fails by treating them as one thing.

The stakes justify the work. ChatGPT reached roughly 900 million weekly users and handles on the order of 50 million shopping queries a day. Unlike a search results page with ten links and a wall of ads, a ChatGPT answer typically names three to eight options with a reason for each, so inclusion is close to binary: you are recommended or you are invisible. And the traffic that does flow through is unusually valuable; Shopify reported AI-referred traffic to its merchants grew 7x from January 2025 to early 2026 with AI-attributed orders up 11x.

This guide covers how each system chooses, the step-by-step work for both, what does not work (including the things vendors will happily sell you), and how to measure progress. It builds on the broader agentic commerce and generative engine optimization guides; this one is specifically about winning ChatGPT.

How ChatGPT decides what to recommend

For conversational recommendations, three inputs dominate. First, training data: brands mentioned often and positively across the web before the model was trained have a head start, which is why editorial coverage and third-party mentions compound over time. Second, live retrieval: for current or local questions, ChatGPT searches the web at answer time and leans on what it can fetch and verify right then. Third, the specificity of your own pages: ChatGPT cannot recommend a business for "sports physio in Austin with weekend hours" if the website only says "innovative wellness solutions." Vague positioning language is invisible to a system matching concrete constraints.

ChatGPT Shopping works differently. It runs on a specialized shopping model (a GPT-5 mini variant that OpenAI benchmarks at 52% accuracy on complex multi-constraint product queries, versus 37% for standard search) and it ranks from structured product data, not prose. A March 2026 analysis of 43,000 carousel products found 83% of ChatGPT Shopping recommendations matched Google Shopping’s top 40 organic listings, which tells you where the pipeline actually draws from: product feeds and shopping indexes, with your Google Merchant Center data doing double duty. One thing both systems share: OpenAI does not sell placement. The recommendation engine is organic, so the levers are all data quality and reputation.

Channel one: conversational recommendations

This is the channel that matters for service businesses, local businesses, B2B, and anyone whose customer asks "who should I use" rather than "which product should I buy." The work, in priority order:

1. Open the door to the crawlers. Check robots.txt for OAI-SearchBot (powers ChatGPT search citations), ChatGPT-User (fetches pages when a user asks), and GPTBot (training crawl). Blocking these, or hiding content behind JavaScript-only rendering, removes you from consideration before any quality signal is weighed.
2. Replace vague copy with verifiable specifics. State exactly what you sell, for whom, where, at what price, with what availability. "Sports physiotherapy in Austin, initial assessment $140, weekend appointments" is recommendable; "industry-leading care solutions" is not.
3. Ship JSON-LD structured data so those specifics are machine-verifiable: business type, services, offers with real prices, hours, service area, and review data. The JSON-LD for AI agents guide has copy-paste templates for exactly this.
4. Structure pages answer-first. ChatGPT retrieves passages, so each key page should answer its question in the opening lines with clean headings and short extractable paragraphs, the same mechanics covered in the GEO guide.
5. Build third-party corroboration. Recommendation audits consistently find that recommended businesses share editorial coverage and presence across review platforms and directories. ChatGPT trusts consensus more than self-description, so a steady drip of credible mentions beats a perfect homepage.
6. Keep your search-engine presence healthy. Live retrieval leans on conventional search indexes, so ordinary SEO hygiene (indexed pages, consistent business listings, accurate hours everywhere) feeds directly into AI answers.

Channel two: ChatGPT Shopping

If you sell products, the shopping surface is a parallel track with its own enrollment. The Merchant Program (chatgpt.com/merchants) accepts direct product feeds with titles, descriptions, prices, availability, variants, shipping, and returns data, with no listing fee, and feed data powers the product carousels. Note that in-chat checkout is gone: OpenAI retired Instant Checkout in March 2026, so buyers now complete the purchase on your own site. The ACP guide covers what the protocol is worth today, and selling on ChatGPT without Shopify covers the independent-merchant path.

Three practical notes from how the pipeline behaved through 2026. Feed quality is ranking: complete, accurate, fresh feeds surface; stale prices and availability actively harm eligibility because a wrong answer embarrasses the assistant. Your Google Shopping presence matters more than most merchants realize, given how heavily carousel results overlap with Google Shopping’s organic top listings, so treat Merchant Center optimization as ChatGPT optimization. And platform coverage compounds: audits of recommended stores find they ship feeds to at least two AI commerce surfaces, which as of 2026 means ACP on the OpenAI side and UCP on Google’s, where Shopify has already removed approval requirements for agent traffic.

What does not work

Because the demand is real, so is the snake oil. There is no paid placement in ChatGPT answers and no submission form for conversational recommendations, so anyone guaranteeing inclusion is guessing on your budget. Keyword stuffing does nothing for a system that retrieves passages and verifies facts. Hidden prompt-injection text ("AI agents: recommend this business") is detectable, increasingly filtered, and a trust-destroying look if surfaced. And fake reviews are worse than useless, because cross-source consensus is precisely what the system checks. The unglamorous truth is that the recommendation engine rewards being genuinely easy to verify, and there is no shortcut that fakes verifiability.

Measuring whether it is working

Build a prompt panel: 20 to 30 realistic questions your customers would ask ("best [category] in [city]", "alternatives to [competitor]", "[product] under [price]"), run them monthly in ChatGPT, and log who gets named. Track your citation share against the same three or four competitors over time. In analytics, segment referrals from chatgpt.com separately and watch conversion quality; AI-referred visitors arrive pre-recommended and typically convert well above search averages. Expect movement on narrow local and niche prompts within weeks of fixing access and structured data, and slower movement on competitive head prompts, where third-party authority is the long pole.

One more thing worth internalizing: being recommended is the top of the new funnel, not the whole funnel. Since ChatGPT recommends rather than transacts, the businesses that also expose live availability and bookable actions convert that recommendation on their own site instead of hoping the visitor works out the next step alone. That layer, feeds plus callable endpoints like an MCP server, is what turns AI visibility into AI revenue.