ChatGPT Ads Are Becoming an Auction

The most important search story this week came from OpenAI. On October 5, it announced visual ads inside ChatGPT and a full measurement stack. Separately, LiveRamp had just expanded its OpenAI partnership into first-party audience activation.

Eight months after ads first appeared in ChatGPT, the product runs at roughly a $1 billion annualized pace, per AIStockWire's August reporting. If your SEO strategy treats AI answers as free organic real estate, the ground just moved.

I tore the announcement apart. Here is what OpenAI is building, and why it matters more to SEO than to paid media.

The five pieces of the October 5 announcement

First, the format. Later this month, OpenAI will start testing visual ads in ChatGPT in the US with a small group of advertisers. The ads appear while a user generates images, clearly labeled and kept separate from the image ChatGPT creates. They are designed to show product inspiration, product use, or experiences tied to a service. Think of someone generating a bedroom mood board and seeing a labeled product tile beside it. If that is how the format renders, it is a new discovery surface sitting between Pinterest and Google Shopping.

Second, the measurement plumbing. Advertisers can now pipe conversion data into ChatGPT Ads from Hightouch, Tealium, and LiveRamp, and connect attribution through AppsFlyer, Adjust, Branch, and Kochava. That addresses one of the three complaints advertisers have raised since the February launch: they could not prove the ads drove sales.

Third, incrementality. OpenAI is working with Haus, Measured, and WorkMagic on geo-based tests. That is the incrementality question most ad platforms dodge for years: did this ad create a sale that would not have happened anyway?

Fourth, brand safety. DoubleVerify and Integral Ad Science are running brand suitability pilots without access to private user conversations. OpenAI says its guardrails keep ads out of sensitive conversations, and qualifying advertisers can block their ads from appearing next to certain phrases.

Fifth, the scale picture. OpenAI says ChatGPT reaches 1.2 billion people a week. For context, Zeno Group's Shamsul Chowdhury told Digiday that ChatGPT handles about 2.5 billion prompts a day against roughly 15 billion Google searches, and that ChatGPT has one ad placement where Google has several. Ad space is the constraint, not audience. Note who sees the ads: Free and Go plans do, while Plus, Pro, Business, Enterprise, Edu, and under-18 accounts do not. The paying audience is partially shielded. The free audience is the inventory.

OpenAI announced it at the start of Advertising Week New York, where its executives went to ask marketers for feedback.

Your customer list, inside ChatGPT

A week earlier, in late September, LiveRamp expanded its OpenAI partnership from measurement into audience activation, live in 11 markets to start. Advertisers can now use RampID to bring first-party data, CRM records, loyalty lists, web and app behavior, into ChatGPT Ads as Custom Audiences for targeting and suppression.

The mechanics will look familiar to anyone who has run Google or Meta ads: upload a CSV or TXT of emails, phone numbers, or hashed versions in OpenAI's Ads Manager, include or exclude audiences at the campaign level, and set bid multipliers from 0.1x to 10x at the ad group level. OpenAI keeps a 25,000 matched-user floor for inclusion targeting, but removed the minimum for exclusion audiences in late August, so suppression lists of any size are usable.

OpenAI made it familiar on purpose: a media buyer who knows Customer Match can operate ChatGPT Ads on day one. Search Engine Journal's reporting adds the caveat: same controls, different performance, and almost no public performance data yet. Suppression is the sensible first test.

The SEO angle

Everyone is fighting to become the cited source inside the AI answer. Meanwhile OpenAI is building the paid placement next to the same answer. Organic citation and sponsored placement will soon compete inside one conversation. I have written about how to earn the citation side. The paid side just got real.

OpenAI's line is direct: advertising does not influence the answers ChatGPT provides. Take that as the current policy until someone independently audits it. Google spent two decades building visible separation between ads and organic results, and the industry still audits that separation constantly. Anyone investing in AI visibility should test OpenAI's claim the way they test everything else: with data.

Branded queries and product questions are where buying intent lives. Those same queries are now answered by the model and monetized by the platform. When a shopper asks ChatGPT which product to buy, the answer can cite your content and show your competitor's ad in the same screen. I covered Google's version of this last week, in Google paying publishers for AI answers. OpenAI's version moves faster and discloses less.

The test I would run

I have not run ChatGPT ads yet, and I will not pretend otherwise. But this is my job, I do search engine marketing for clients at First Rank, so here is the test I would design for a client asking about it this month. The early consensus matches how I would design it anyway: SEJ's reporting calls suppression the sensible first test, and OpenAI's own incrementality partners run geo-holdouts. So: customer list, exclude recent buyers, one category, geo-holdout. Compare attributed cost per acquisition against the blended paid search benchmark. Treat the early vendor numbers as directional: Rockerbox reported one advertiser, WeightWatchers, saw attributed CPA 15.3% below its paid search benchmark, and Triple Whale reported 93% of one retailer's ChatGPT-referred visitors were new. Two data points from vendors with a stake in the outcome. They warrant a test.

The test must answer whether conversation context converts differently than search intent. Someone asking ChatGPT to "help me pick a CRM for my 20-person agency" and someone typing "best CRM for a 20-person agency" into Google look like the same buyer. They may not buy the same way, and they will not cost the same to reach. My framework for any channel is intent times acquisition cost times conversion rate times margin: small test, real measurement, kill it fast if the math fails.

What this means for a store like mine

I also run an e-commerce store (StickyBra, linked from my homepage), and the visual ad format is the piece I am watching closest. My products are things people need to visualize: fit, shade, how they look under clothing. An ad that appears beside a generated image, showing product inspiration at the exact moment someone is imagining an outfit, is a placement built for categories like mine.

I would watch this placement before I buy it. The US-only pilot, the single ad slot, and zero public pricing benchmarks mean early CPCs could go anywhere. Watching is the position until the data exists.

The bottom line

Citations earn you a mention inside the answer. Ads buy you a seat next to it. The brands that win AI search will hold both, and the ones who only do SEO will discover the answer box has an auction they never bid in.

Are you treating ChatGPT as a channel to measure, or still as a curiosity to watch?