ChatGPT Ads Are Live.
This Is Your Window.

Paid advertising has arrived inside the AI assistants people now ask what to buy. ChatGPT runs a self-serve ad platform with shopping cards and a conversion pixel. Google serves ads inside AI Overviews and AI Mode. Both are early, cheap, and measurable. The play is the same on any new channel: measure what AI already says about a brand, fix the free half, then buy the gaps with attribution that proves what each click did. SearchShopAI runs that whole loop for beauty and ecommerce brands.

• The window is open nowChatGPT ad penetration jumped from 0.03% of all responses to 47.7% of US responses in three weeks (peaking on 26 May 2026). The auction is thin and the competition has not arrived yet.
• Cheap and self-serveThe Ads Manager went live to every US business on May 5, 2026 with no minimum spend, clicks around $3 to $5, and CPMs that have fallen to $25 to $45.
• Measurable from day oneOpenAI ships a browser pixel and a server-side Conversions API. Google folds AI placements into existing Search, Shopping and Performance Max campaigns.
• Organic first, then paidThe product feed that fixes a brand's organic AI answers is the same feed that powers its shopping ads. Fix the free half, then buy the gaps.

What happened

OpenAI moved fast. A US ad pilot opened in February 2026, expanded to Canada, Australia and New Zealand in March, and on May 5 a self-serve Ads Manager went live to every US business with no minimum spend and clicks around $3 to $5. Then it scaled. An independent Writesonic study of 363,129 ChatGPT responses caught ad penetration jump from 0.03% to 14.12% overall, and to 47.7% on US responses, on May 26, a 280x increase that held the next day. By one count, more than 1,400 brands had run ChatGPT ads in the US between February and late May (Sensor Tower, US Android app).

Best Buy and beauty brand e.l.f. were among the advertisers the Writesonic study observed. eMarketer separately reported retail and grocery as the single largest ad category, about 44% of ChatGPT ad impressions in a February 12 to 26 window, with brands including Ulta and Sephora appearing among them. The UK went live June 6. Mexico, Brazil, Japan and South Korea are next.

Writesonic sells AI-visibility tracking, so read its penetration figures as a vendor's. A second public study published the same week landed within one percentage point of the US number, which is the reason to trust the order of magnitude here.

What a ChatGPT ad actually looks like

Ads appear below a completed answer, clearly labeled Sponsored, and do not change the answer above them. For a beauty brand, that means a product can appear right where a shopper asks ChatGPT for the best routine for hormonal acne or what a vitamin C serum actually does. The formats are built for commerce.

• Sponsored resultThe baseline unit: a single sponsored card under the response with a short headline, image, favicon and link. As of May, it can carry a personalizable call-to-action button: shop now, book now, sign up, learn more.
• Shopping cardA dedicated e-commerce format that shows price and customer reviews, in portrait or landscape. It pulls live product data, so accurate feeds and reviews are the lever.
• Stackable carouselPortrait shopping cards stack so three to four ads sit side by side, a carousel right where a shopper is already asking which product to buy. Built for product-comparison moments.
• Catalog-driven adsUpload a product feed (Google Shopping format, up to a million SKUs per advertiser) and ChatGPT auto-generates ads from the catalog. The same feed work that fixes organic accuracy powers the ads.

Targeting is contextual to the conversation, not keywords. Ads show to logged-in adults on the Free and Go tiers, roughly 85% of users. Plus, Pro and Enterprise subscribers see none.

Google's AI ads are live too

ChatGPT built a new ad surface. Google did the opposite. It folded AI into the campaigns advertisers already run. Text and Shopping ads from existing Search, Shopping and Performance Max campaigns are now eligible to serve inside AI Overviews (above, below, or within the AI answer) and inside AI Mode, Google's conversational search. Both are labeled Sponsored. As of mid-2026, Google lists ads in AI Overviews as live in 12 English-language markets (the US, Canada, Australia, New Zealand, India, Indonesia and others), not everywhere yet.

The mechanism is the inverse of opt-in. There is no separate placement to buy or opt out of. Eligibility flows automatically from AI-friendly setups: broad match, AI Max for Search, Performance Max. In AI Mode, Gemini writes the ad creative per query, so structured product data and landing-page quality become the performance levers, not static assets. One clarification the coverage keeps getting wrong: the Gemini app itself has no ads. AI Mode is Gemini-powered search, and that is where the ads are.

Why early matters

• Cheap because the floor droppedThe honest version: prices are low because OpenAI cut the spend floor from a six-figure minimum (around $200K at the February pilot) to $0, not because a mature auction is bidding clicks up. CPMs fell from about $60 to $25 to $45 in nine weeks. The competition has not arrived yet.
• The same pattern as Google in 2002Trade press draws the parallel directly: early Google Ads and early Facebook buyers got favorable rates, then prices climbed as the platforms matured. This is the pre-global-auction moment, and it will not last.
• Measurable from day oneOpenAI ships a browser pixel and a server-side Conversions API with a click identifier and hashed-email matching. A team can install it, then layer first-party attribution on top, so the report shows revenue per platform instead of a guess.
• Worth saying plainlyIt is early. OpenAI's own reporting is aggregate-only and unverified, and some shoppers distrust ads in AI answers. That is exactly why a measured, organic-first approach wins here, and why it is best treated as a learning bet, not a blank check.

How you will know it worked

The hardest part of a new ad channel is proving it paid off. OpenAI gives advertisers aggregate numbers, impressions and clicks, and nothing else. Closing that loop is the whole point.

SearchShopAI runs a first-party attribution rail for organic AI traffic: a store pixel, signed cart attribution that survives to the order, and a closed-loop join that ties a click to the actual purchase. That rail extends to paid. The OpenAI pixel and Conversions API get installed, the rail layers on top, and the report shows revenue per platform, deduped and server-side, not just a click count taken on faith.

One model for the whole AI surface, organic and paid, so a brand can compare what a ChatGPT click is worth against a Google or Meta one. See how the integration works.

What to do, in order

• 1. Measure what AI already says about the brandBefore spending a dollar, get the organic baseline: which products ChatGPT, Gemini, Copilot and Google AI recommend, what they get wrong, where the purchase path breaks. That is exactly what the SearchShopAI free audit reports, across five platforms.
• 2. Fix the free halfFeed, schema, crawler access, purchase path. The product feed that fixes the organic answers is the same feed that powers the shopping ads, so this work pays twice. It lands the same week and keeps paying after the ad budget stops. The full playbook is here.
• 3. Instrument before spendingInstall the OpenAI pixel and Conversions API, the Google Ads tag, and clean UTM conventions before the first click. Conversion-optimized bidding needs a conversion signal flowing first. No measurement, no optimization, no idea what worked.
• 4. Buy the gaps, not the brandStart where organic takes months: non-branded category queries (best vitamin C serum), competitor comparisons, new launches. On Google, get AI-friendly (broad match, PMax, AI Max) to stay eligible for AI Overviews and AI Mode. Begin small, around $20 to $50 a day, and read the data.
• 5. Manage it like performance mediaTest creative (CTA buttons and shopping cards beat plain text), watch CPA not just clicks, and treat the whole AI surface, organic plus paid, as one funnel with one attribution model. This is the part most brands get wrong, and the part a performance team handles for a living.

SearchShopAI runs AI ad campaigns with the attribution to prove what worked. Ask about paid AI ads management on a demo call.

Why trust SearchShopAI with the spend

This is not a hand-off to a junior media buyer learning a new button. SearchShopAI's paid media is led by operators who have run performance marketing at the highest level for over a decade.

In previous roles, the team has managed acquisition and retention for some of the largest consumer brands in streaming, gaming and telecom, including Netflix, Spotify, Blizzard, AT&T and YouTube TV, with monthly spend into eight figures (up to $13M) and the ROAS to answer for it. The team runs every surface that matters for beauty: Meta Advantage+, TikTok Spark and Shop ads, Google PMax, Shopping and Demand Gen, Pinterest, Reddit and The Trade Desk, and now the AI-native ones (ChatGPT Ads, Google AI Max, and ads in AI Overviews and AI Mode).

That work came with real measurement rigor: media mix modeling, incrementality testing, multi-touch attribution, and surviving every platform shift from SKAN to the death of the third-party cookie. AI ad inventory today is where mobile was in 2012 and search was in 2002, cheap and under-contested for whoever moves first. SearchShopAI has run the early-channel playbook before, and now points it at the newest surface, with the attribution to prove it worked. If you would buy a click here, talk to SearchShopAI before you do.

Where AI ads go next

Paid AI ads are barely six months old, so the playbook above is the present, not the ceiling. A few shifts are already visible at the edges.

• Landing pages generated per query and pathThe destination stops being one static page. It gets assembled from the exact conversational query and where the shopper sits in the journey, so a question about sensitive-skin retinol lands on a page that answers that, not a generic catalog page. Google already automates parts of this with auto-created assets and final-URL expansion.
• The ad becomes the checkoutAs agentic-commerce standards (the Agentic Commerce Protocol, Google's AP2) make signed in-conversation purchase routine, a sponsored result carries a buy action that completes inside the chat instead of a link back to a site. The click disappears and the unit becomes a completed order.
• Bidding moves from keywords to intentConversational surfaces price the meaning of a request and the shopper's context, and creative gets generated and tested per shopper against the answer flow.
• The surface fragments across assistantsGemini, Copilot and Perplexity are building ad inventory of their own, so one keyword list gives way to one product feed and one set of claims every assistant can read the same way.

These surfaces are small and new, which makes them easy to dismiss. Every shift before this one, search, then mobile, then social, paid off most for the brands that showed up while inventory was cheap and the rules were still loose. AI assistants are fast becoming where buying decisions happen, and these ad formats are how brands reach them. What looks optional today will be a primary way customers are won, and the work of showing up well starts now.