// HOW-TO · AI COMMERCE

How to sell on ChatGPT: driving revenue through ads and GEO (2026)

Sell on ChatGPT in 2026: get your products named in buying answers, run ads to capture demand, and convert the click on your own site.

Last verified · 2026-09-17 · by Moe Ameen

Selling on ChatGPT is not one feature you switch on — it is a buying journey that starts inside the assistant and, in 2026, finishes on your own site. A shopper asks the model for the best product for their problem, the model names a shortlist, sometimes a sponsored unit sits below the answer, and the buyer clicks through to purchase. OpenAI experimented with in-chat purchasing through its Agentic Commerce Protocol, then pulled back native checkout and repositioned ChatGPT around product discovery, so the durable model this year is blunt: discover in AI, buy on your own site. That changes where the revenue work lives — you win the mention and the click inside ChatGPT, then you have to close the sale on a page you control.

This walks the selling funnel end to end, in order. You learn the three ways a sale can start in a ChatGPT answer, baseline how the model sees your products today, make your catalog discoverable, earn the organic recommendation for buying-intent prompts, run ads to capture the high-intent moment, wire the click to convert on your site, and measure revenue rather than visibility. The through-line: presence in the answer is worthless if the click lands on a page that doesn't finish the conversation, so this is as much a conversion job as a visibility one. For the deeper channel strategy behind the two answer slots, see [how to build a ChatGPT ads and GEO content strategy](/how-to/build-a-chatgpt-ads-and-geo-content-strategy).

The steps

Know the three ways a sale can start in a ChatGPT answer. There are three distinct surfaces, and they behave differently. The organic product mention is the model naming your product in the answer itself when someone asks what to buy — earned through generative-engine optimization, no platform fee. The paid slot is the sponsored unit that can appear below the answer, which you bid on in OpenAI's Ads Manager. Product discovery is the shopping-style surfacing of items, which OpenAI now leans on after walking back in-chat checkout. On all three, the purchase itself completes on your site, so treat each as a way to earn a high-intent click, not an in-chat cash register.
Baseline how ChatGPT sees your products today. List the real buying-intent prompts in your niche — 'best [product] for [use case]', 'is [brand] worth it', 'what should I buy to [outcome]' — and run each one. Record three things per answer: whether the model names your product organically, whether a sponsored unit shows for that topic in your market, and whether your items surface at all in any shopping view. This prompt set is both your competitive map (which brand owns each buying answer today) and, later, your scorecard. If you are absent from every answer, that gap is the whole opportunity.
Make your catalog legible to the model. The organic and discovery surfaces reward clean, structured, trustworthy product information — accurate titles, clear specs, current pricing and availability, and genuine reviews, published on pages the model can actually crawl. Allow OAI-SearchBot, the retrieval crawler ChatGPT uses to read the live web; a blanket AI-bot block removes you from its answers entirely. This is the no-fee lever: unlike the paid slot, discoverability is bought with feed and content quality, not budget, so a well-described catalog is the cheapest revenue work you can do.
Earn the organic recommendation for buying-intent prompts. Being crawlable gets you read; being the cleanest answer gets you named. For each priority prompt, publish a genuinely useful buying guide or comparison on your own site — a direct-answer passage the model can lift, backed by real specs, honest trade-offs, and a named source or original data. Then corroborate it: the same accurate product claims echoed across the trusted surfaces the model reads. Rising organic mention share is the compounding, unbuyable half of selling here, so start it before you spend a dollar on ads.
Run ads to capture the high-intent moment. Once the account is set up, ads let you buy placement in buying conversations you don't yet win organically. The native unit is restrained — brand, headline, short description, image, link — so write it like a search ad, not a social one, and build a set of creatives to test against the same intent rather than a single asset. Targeting is contextual: the topics you supply are hints, and the model decides whether your ad fits the actual chat. Ads show only to logged-in users on eligible tiers and rolled-out markets, so treat the paid slot as demand capture on top of the organic reach, not a replacement for it.
Wire the click to convert on your own site. This is the step most brands skip, and it is where the revenue is won or lost in 2026. Because checkout has moved back to the merchant, the click from ChatGPT lands on your site — so send it to a page built for the exact question the shopper asked, not a generic homepage. Continue the thread: the same product, the same problem, the specs and proof the answer implied, and a fast path to buy. Add UTM tags and conversion tracking so you can attribute revenue to ChatGPT-sourced clicks; a high-intent visitor dropped on the wrong page converts worse than a cold one.
Measure revenue, not visibility. Visibility is a leading indicator; sales are the scorecard. Attribute revenue and new customers to ChatGPT-sourced sessions using your tagged links and analytics, and re-run your prompt set on a schedule to log where the model names you and where an ad shows. Read them together per prompt: are you present in the answer, is that presence lifting the conversion rate on the paid click, and is ChatGPT bringing a higher new-customer rate than your other channels? Early merchant data suggests AI assistants can drive strong new-customer acquisition even when in-chat conversion lags — which is exactly why closing the sale on a strong owned page matters.
Sustain it — recommendations and creative both decay. Neither half is set-and-forget. Ad creative fatigues in the auction, so keep feeding fresh on-brand units and retiring tired ones. The organic recommendation is a maintained asset: the index re-crawls on a lag, competitors publish newer guides, and prices and stock drift, so keep product data, buying guides, and last-updated dates current, and add pages as new buying prompts emerge. Re-check the prompt set monthly and judge movement over weeks, not days — ChatGPT's index lags, so give edits time before you kill them.

Common gotchas

Expecting an in-chat checkout to carry the sale. OpenAI walked back native purchasing and repositioned around discovery, so the durable 2026 model is discover in AI, buy on your own site — build the owned page that closes the click, don't wait for a cash register inside the chat.
Buying ads while a competitor owns the organic answer above them. The recommendation the shopper already trusts sits above your ad; if the model names a rival there, your paid click is arguing uphill. GEO is what makes the ad convert, not a separate project you can defer.
Blocking OAI-SearchBot with a blanket AI-bot rule. It removes your products from ChatGPT's organic and discovery answers entirely, and it's invisible until you notice you're never named. The retrieval crawler is separate from the training crawler.
Sending the click to a generic homepage. A ChatGPT click is high-intent and often the end of a long research thread — dump it on a homepage and you lose the sale a competitor's focused landing page would have closed.
Treating ad topics as exact-match targeting. OpenAI is explicit that the topics you supply are hints; the model decides placement from the real conversation. A broad, generic buy gets filtered out — sharp relevance to a defined buying intent earns the slot.
Describing your product differently in the ad and on your pages. Inconsistent names, specs, or prices between the ad and the cited content suppress citations and confuse the model about what you actually sell.
Judging the channel on visibility alone. Being named feels like winning, but the only number that pays is attributed revenue and new customers from ChatGPT-sourced clicks — track the sale, not the mention.

Legal note

You cannot pay OpenAI to change what ChatGPT recommends organically — OpenAI states the ad system is separate from the chat model, so a sponsored unit does not buy the recommendation above it; do not imply otherwise to customers or clients. Ad creative must comply with OpenAI's advertising policies, and every product claim in either the ad or the cited content must be truthful and substantiated — the usual advertising and consumer-protection rules apply in full. Ads are shown only to logged-in adults on eligible tiers and markets, targeting is contextual, and advertisers do not receive users' chats, names, emails, or IP addresses. Commerce features, fees, and availability have shifted during 2026; confirm current merchant requirements and any transaction terms against OpenAI's own documentation before relying on them.

Where Kompozy fits

The 2026 selling model puts the burden in a specific place: the sale completes on your own site, so the content that earns the mention and the content that closes the click are both yours to produce — a buying guide the model can cite, a comparison page a shopper lands on, product stills that read as premium, and the steady social-and-email presence that keeps the model's picture of your brand current. That is a lot of owned content, produced continuously, on one identity. It is a supply problem, and it is the exact job [Kompozy](/) is built for.

Kompozy is a full AI content generation and multi-platform publishing engine, not a repurposing tool, and here it stocks the whole selling funnel from one brand identity. For the organic recommendation (steps 3–4), it generates the citable footprint the model reads and the conversion pages the click lands on: Blog Articles structured as direct-answer buying guides and comparisons for your own site — which are simultaneously your step-6 landing pages — plus Text Posts, brand-exact Carousels, and Email Newsletters that echo the same product claims across the surfaces ChatGPT crawls. For the paid slot (step 5), one product becomes a test set of on-brand stills — Photo Posts, Infographic Photos, and Persona Photos — so 'build a set, not one asset' is picking a winner in review instead of a new shoot. A single [Persona Brief](/glossary/persona-brief) holds your product name, specs framing, and voice identical across the ad and the citation, which is exactly the consistency the gotcha about 'same product, described the same way' demands, and the per-post review gate rejects invented numbers before they ship, so the claim the model lifts from you is a real one.

Be exact on the boundary: Kompozy produces and publishes the content the funnel runs on and builds the owned pages that convert the click — it does not bid in OpenAI's Ads Manager, run your product feed or catalog, process the payment, or track your revenue attribution. You own the store, the ad account, and the analytics; Kompozy owns the supply that gets you named and the pages that close the sale. [Autopilot](/glossary/autopilot) then keeps that footprint fresh across the eight social platforms plus blog and email from one queue, on the monthly cadence step 8 demands. Creator ($49/mo for 2,500 credits) fits a solo brand building buying guides and a creative batch around one product line; Pro ($299/mo for 18,000 credits) sustains a team feeding both slots across a growing catalog; Enterprise is custom for agencies running the channel across many brands.

Frequently asked questions

Can people buy directly inside ChatGPT?

OpenAI launched in-chat purchasing (Instant Checkout, built on its Agentic Commerce Protocol) in September 2025, then pulled it back in March 2026 and repositioned ChatGPT around product discovery instead. The durable model this year is 'discover in AI, buy on your own site': the assistant helps a shopper research and decide, then the purchase completes on the merchant's own store. So the practical goal is to win the mention and the click, then convert on a page you control.

How do I get my products recommended by ChatGPT?

Make your catalog crawlable and legible — accurate product info, clear specs, current pricing, and genuine reviews on pages the model can read, with OAI-SearchBot allowed. Then publish the cleanest buying guide or comparison on the web for each buying-intent prompt, backed by real specs and honest trade-offs, and corroborate the same claims across the trusted surfaces the model reads. Organic recommendation is earned with content and feed quality, not paid for.

Do I need ChatGPT ads to sell, or is GEO enough?

They do different jobs on the same buying moment. GEO earns the more-trusted, free, but slow and unbuyable organic mention; ads buy a labeled, controllable placement to a slice of the audience, fast. Their weaknesses are opposite, so running both covers the whole answer, and an ad aimed at a conversation where you're already recommended converts far better than a cold one. Start GEO first, then layer ads onto the prompts where your organic presence is strongest.

Where does the sale actually happen?

On your own site, in the 2026 model. After OpenAI walked back in-chat checkout, the click from an organic mention, an ad, or a discovery surface sends the shopper to your store to complete the purchase. That makes your landing page part of the selling machine, not an afterthought — the page has to continue the exact conversation the shopper was in and give them a fast path to buy, or the high-intent click is wasted.

How do I measure whether ChatGPT is driving revenue?

Tag every ChatGPT-destined link with UTMs and read attributed revenue and new customers in your analytics, not just impressions. Re-run your buying-intent prompt set monthly to log where the model names you and where an ad shows, and read the two together: is your presence in the answer lifting the conversion rate on the click, and is ChatGPT delivering a higher new-customer rate than other channels? Attributed sales, not mentions, are the scorecard.