What We Know About Paid Ads in AI: A Practical, Evidence-Based Guide for 2026

Paid ads in AI are here. They are already shaping how customers find brands inside ChatGPT, Google AI Overviews, and Microsoft Copilot. The pilot data is messy, the pricing keeps changing, and the reporting tools are early. But there is enough verifiable information now to plan around, and the brands that wait for the picture to clear are likely to find competitors and affiliate publishers already occupying their placements.

This guide covers what is publicly verifiable about paid AI ads today, the risks that have shown up in the first months of live testing, and where Answer Engine Optimization (AEO) fits into the picture. One pattern keeps repeating across the data. Paid AI works better when the AI already understands and trusts the brand placing the ad. The work that builds that understanding is AEO.

Key takeaways

• OpenAI's ChatGPT ads pilot has been live since February 9, 2026 for U.S. Free and Go users. Plus, Pro, Business, Enterprise, and Education tiers stay ad-free. The pilot is now expanding to Canada, Australia, and New Zealand.
• Google already serves ads above, below, and inside AI Overviews in 200-plus markets, with ads in AI Mode and sponsored retailer placements in active testing.
• Anthropic took the opposite bet. Claude will stay ad-free, a position Anthropic put on national television during the Super Bowl.
• ChatGPT ads cluster on commercial evaluation prompts. About 61% trigger on "best" queries, 31% on "X vs Y" comparisons, and only 9.5% on how-to questions.
• Brand hijacking is already common. On "X vs Y" comparison prompts, 86% of the ads served belong to brands that were not even in the comparison.
• Citation in AI answers correlates with stronger paid performance. Seer Interactive found brands cited in Google AI Overviews earn 91% more paid clicks and 35% more organic clicks than uncited brands.
• Instant Checkout in ChatGPT is now live via Stripe's Agentic Commerce Protocol. Product ranking inside that flow is driven by product data quality, not ad spend.
• AEO is the layer that decides how every future AI surface treats your brand. Paid AI works better when it lands on top of a strong AEO foundation.

Paid ads in AI are already here

OpenAI confirmed in a January 16, 2026 announcement that it would begin testing ads in ChatGPT for adult Free and Go users in the United States. Plus, Pro, Business, and Enterprise tiers stay ad-free. Testing officially started on February 9, 2026, reaching about 1% of mobile users initially and expanding to roughly 5% by late March. Early ad pricing was reported at approximately $60 CPM, with a cost-per-click model added in April carrying advertiser bids in the $3 to $5 range.

The ads themselves sit at the bottom of ChatGPT responses, labeled "sponsored," and are visually separated from the AI's answer. According to OpenAI, ad targeting works against "the topic of your conversation, your past chats, and any previous ad interactions." The OpenAI announcement confirmed plans to expand the pilot to Canada, Australia, and New Zealand in early 2026, with more markets to follow.

Google has been further along, for longer. The official Google Ads documentation confirms that ads are eligible to show above or below AI Overviews in all 200-plus markets where AI Overviews are available, and that ads inside AI Overviews are now available in English on mobile and desktop in countries including the United States, Canada, India, and Australia. These placements draw from existing Search, Shopping, and Performance Max campaigns, and Google says matching is based on both the user's query and the actual content of the AI Overview. At Google Marketing Live 2025, Google said it was expanding ads in AI Overviews and testing ads integrated into and below AI Mode responses, plus clearly labeled sponsored retailer placements and "Direct Offers" in AI Mode.

Gemini itself is the open question. In late January 2026, Google DeepMind CEO Demis Hassabis said there were "no plans" for ads in Gemini, per NBC News. That statement aside, Google's existing ads infrastructure, advertiser base, and revenue model still point to ads in Gemini being a likely 2026 development, especially as Gemini gets more deeply embedded in Workspace, Docs, Gmail, and Drive.

Microsoft has been publishing internal numbers from Copilot. Its advertising team reports that Copilot users generated 73% higher click-through rates and 16% stronger conversion rates than traditional search in its own data, with a 33% shorter customer journey. Useful directional signal, but it is vendor-reported data, not a neutral benchmark.

Among the major frontier labs, Anthropic took the opposite position. On February 4, 2026, the company stated, "There are many good places for advertising. A conversation with Claude is not one of them," and added, "Claude will remain ad-free." Anthropic then ran four Super Bowl commercials with the tagline "Ads are coming to AI. But not to Claude." Perplexity also stepped back from sponsored answer experiments earlier in 2026, citing concerns that paid placements would erode user trust.

These competing positions matter because the format is still being negotiated in public. The platforms that keep ads useful, clearly labeled, and separated from the underlying answer are the ones most likely to keep their users with them.

The Practical Reality of Conversational AI Advertising

Advertisements within ChatGPT appear at the bottom of the chat interface, clearly separated from the AI's generated response and labeled as sponsored. To preserve privacy, OpenAI does not share raw user conversations with advertisers. Instead, the system matches ads to users contextually, analyzing the current topic, past chats, and prior interactions with advertisements.

Major brands have joined the early pilot tests. Adobe partnered with OpenAI to run conversational ads for Acrobat Studio and Adobe Firefly through the marketing agency WPP Plc. However, the initial entry barriers for these tests were steep, with media buyers reporting a minimum budget commitment of 200,000 dollars and a cost per thousand views, or CPM, of roughly 60 dollars.

To attract performance-driven marketers, OpenAI has introduced a cost-per-click, or CPC, bidding model, allowing pilot advertisers to set bids between three and five dollars per click. Even with this shift, conversational ad tracking is still highly immature. Click-through rates often fall below traditional Google search benchmarks, and advertisers struggle to link clicks directly to sales.

A study by Adthena found only 580 total ads in the ChatGPT ad index. Due to uncertain returns, 36 percent of observed brands ran an ad only once and did not return, executing a test-and-pause strategy.

Meanwhile, Google is preparing to deploy Google Ads inside Gemini by 2026, building on its existing sponsored placements in Google AI Overviews and AI Mode. Conversational ads on Google require brands to surrender control, trusting Google's machine learning systems to automatically modify and tailor ad copy to fit long, highly descriptive conversational queries.

The following table compares the current status of advertising across the primary generative AI networks.

The strongest public evidence so far

‍The audience is now mainstream

‍OpenAI says more than 700 million people use ChatGPT every week. Google reported in May 2025 that AI Overviews had reached 1.5 billion monthly users across 200 countries and were driving more than 10% growth in usage for the queries that show them in major markets. ChatGPT alone recorded 5.73 billion visits in March 2026 according to data cited by PPC Land.

Commerce signals are picking up speed

‍Adobe reported in August 2025 that traffic from generative AI sources to U.S. retail sites was up 4,700% year over year in July 2025, and 38% of consumers surveyed said they had already used generative AI to shop online. Bain reported that ChatGPT shopping prompts roughly doubled in popularity between January and June 2025.

Where ads actually trigger

‍Adthena CMO Ashley Fletcher has published the most detailed live analysis of ChatGPT ads since launch. From a sample of 580 ChatGPT ads pulled in April 2026, Fletcher found that 61% of ads triggered on prompts beginning with the word "best," 31% triggered on "X vs Y" comparisons, and only 9.5% triggered on how-to questions. He described that intent profile as "almost identical to Google Shopping, bottom-funnel, comparison-led, transactional." ChatGPT ads are landing primarily on commercial evaluation queries, not broad discovery.

How ads behave in the chat itself

‍Adthena's analysis of 1,500-plus prompts in the first weeks of the pilot found that ads appear as branded cards embedded directly in the conversation flow, carry a "Sponsored" label with a brand favicon, and are matched purely to the user's question rather than to cookies or demographic profiles. As Adthena summarized, "Ads on ChatGPT are matched to what a user asks, not who they are." That is a meaningful difference from traditional programmatic and paid social.

AI surfaces compress organic clicks, but citation buys back performance

‍Ahrefs analyzed roughly 300,000 keywords and found that AI Overviews reduced click-through rate to the number one organic result by about 34.5% in its initial study, with later updates showing larger declines. Seer Interactive analyzed 3,119 informational queries across 42 organizations and reported a 61% drop in organic CTR and a 68% drop in paid CTR for queries that triggered AI Overviews. The same study found that when a brand was cited inside the AI Overview, it received 35% more organic clicks and 91% more paid clicks compared to uncited brands.

Seer's lead researcher Tracy McDonald is careful to qualify the finding. "We cannot definitively prove that citation causes higher CTRs. It is equally possible that brands with stronger authority and higher baseline CTRs are simply more likely to be cited by Google's AI. What we can say with confidence is that queries where you are cited consistently outperform those where you are not, regardless of the causal direction." Seer's April 2026 update also found that organic CTR on AI Overview SERPs rebounded 85% in the first two months of 2026, climbing from 1.3% in December 2025 to 2.4% in February 2026. The earlier assumption of permanent collapse looks premature, though CTRs have not returned to pre-AIO levels.

Taken together, these are the cleanest signals available today. The audience is huge, the ad inventory is real, the intent profile is bottom-funnel, the ad mechanics are working, and brand citation in AI answers consistently outperforms non-citation on both organic and paid sides.

New risks paid search did not have

Two risks have already surfaced in the pilot data that businesses need to plan around now, not later. A third one is about measurement.

1. ‍Brand hijacking on comparison prompts. Adthena's analysis of 177 comparison prompts found that when users typed "X vs Y," the ad shown was not one of the named brands in 86% of cases. Only 14% of those placements actually belonged to the brands being compared. Examples surfaced by Adthena include Dollar General ads on prompts comparing Drizly and DoorDash, and HubSpot ads on prompts comparing Zapier and Make. If you sell into a competitive category and you do not defend your own comparison prompts, a competitor or an unrelated brand may be the only sponsored result a buyer ever sees.‍
2. Affiliate publishers claiming commercial space. Affiliate publishers such as WSJ Buy Side and Forbes moved quickly into ChatGPT ad inventory, occupying a meaningful share of commercial placements on "best [category]" queries. They run paid placements to send users to their own review pages and capture affiliate commissions, with the actual brands often absent from the answer. This is not new to AI, but it concentrates faster in a conversational interface where there are fewer placements per response than in a traditional SERP.‍
3. Measurement is still immature. OpenAI's Ads Manager currently reports impressions, clicks, spend, CTR, average CPC, average CPM, and conversions, and supports static URL parameters such as UTMs on landing pages. Google does not let advertisers isolate AI Overview placements cleanly. They run inside Search, Shopping, and Performance Max campaigns and are reported as Top Ads. A clean read on AI-driven traffic right now requires your own UTM conventions, dedicated landing-page tracking, and disciplined post-click reporting. Treat AI-source traffic as its own segment in analytics from day one.

Agentic commerce is quietly redrawing the floor plan

Ads are only one part of the monetization story in AI. The other is the checkout itself, and it is moving faster than the ad market.

On September 29, 2025, OpenAI and Stripe announced Instant Checkout in ChatGPT, powered by a new open standard called the Agentic Commerce Protocol (ACP). U.S. ChatGPT Plus, Pro, and Free users can buy directly from U.S. Etsy sellers in the chat, with over one million Shopify merchants (including Glossier, SKIMS, Spanx, and Vuori) onboarding soon after. PayPal's ACP server is set to bring tens of millions of small businesses onto the platform in 2026.

OpenAI is explicit that Instant Checkout is not pay-to-rank. From OpenAI's announcement: "Merchants pay a small fee on completed purchases, but the service is free for users, doesn't affect their prices, and doesn't influence ChatGPT's product results." Product ranking is based on factors like availability, price, quality, whether a merchant is the primary seller, and whether the merchant has Instant Checkout enabled. Fidji Simo, OpenAI's CEO of Applications, said the goal was to make it "possible for businesses of all sizes to meet people where they are, and for shoppers to complete purchases seamlessly in conversation."

By early 2026, Shopify had rolled out Agentic Storefronts that syndicate merchant catalogs to ChatGPT, Microsoft Copilot, and Perplexity at once, with AI channel attribution exposed in the Shopify admin. What this means for any business that sells physical products is simple. Ads are half the conversation. The other half is whether the answer engine can find your product in the first place. Product feeds, accurate GTINs, descriptive titles, real-time availability signals, and structured product data are quickly becoming the front door to AI commerce, separate from any ad spend.

Why AEO becomes more important, not less

A common assumption is that paid ads in AI will replace AEO. The current evidence points the other way.

Google's own documentation says ads in AI Overviews are matched against both the user query and the content of the AI Overview itself. OpenAI's targeting works against the topic of the conversation, the user's recent context, and prior ad interactions. Seer's published data shows cited brands outperform uncited brands on both organic and paid CTR by a wide margin. The same signals that decide what gets cited in the answer also decide what kind of frame surrounds the ad. AI retrieval systems are using the markers a careful human editor would use to decide what is worth quoting: clear structure, named experts, specific statistics, authoritative sources, precise terminology.

Google's own guidance on optimizing for generative AI features confirms the same picture from the platform side. The signals AI systems use are the fundamentals that have always defined good SEO: unique, non-commodity content; clear structure; crawlability; freshness; accurate business and product information; content that is easy to retrieve and quote. The work has not changed in principle. It has become visible in a new surface area, and the rewards for doing it well are now showing up in places they did not show up before.

That is what a strong AEO program is built for. AEO is the discipline of making your content, product data, and brand signals legible to AI systems so they can retrieve, trust, and cite you. When that foundation is in place, paid ads land in a context the AI has already prepared. When it is not, paid ads still run, but they land in a less favorable frame, often next to a competitor or an affiliate publisher that did do the work.

Paid AI ads work better when the AI already understands the brand. AEO is what builds that understanding. The two layers compound.

What smart businesses should do right now

A practical playbook for the next 12 months, based on what is verifiable today:

• Defend your brand comparison queries first. Run a monthly audit of "X vs Y" prompts that include your brand. With 86% of those placements currently going to unrelated brands or competitors, buying coverage on your own comparison terms has gone from nice-to-have to baseline.
• Treat your product feed as media infrastructure. Google's Merchant Center product data powers Shopping ads, AI Max, AI Mode, and AI Overviews. OpenAI's merchant documentation says structured product feeds make products discoverable in ChatGPT and improve accuracy and freshness. If you sell products, your feed is part of your acquisition engine.
• Move past keyword-only targeting in paid. Google states that ads in AI Overviews require AI-powered targeting such as broad match, AI Max, Performance Max, or Shopping. Accounts built only around tight exact-match logic are under-covered in AI surfaces.
• Build pages AI can quote. Lead every important page with a clear, self-contained answer in the first 200 to 300 words. Use H2 and H3 structure so retrieval systems can extract the right chunk. Cite authoritative sources, include precise statistics, and quote named experts with their titles and affiliations. The published research lines up on this point: those signals raise citation likelihood.
• Layer in your own measurement. Until platform reporting matures, the cleanest read on AI-driven traffic comes from disciplined UTM conventions, dedicated landing pages, and post-click reporting in your analytics platform. Track AI-source traffic as its own segment from day one.
• Keep disclosures and claims tight. The FTC's .com disclosure guidance says required disclosures need to be clear, conspicuous, close to the claim, and understandable. In a conversational interface where ads sit beside a trusted assistant's answer, trust is a performance variable.
• Run AEO and paid AI as one program. They share inputs, they share targets, and they share the brand's job to be done in the AI's answer. Coordinating them is where compounding returns come from.

What we still do not know

There is more legitimate uncertainty in paid AI than the loudest takes suggest.

Public, independent ROAS and CPC benchmarks for AI-native ad units are still scarce. Most hard performance claims today come from platform-owned studies, vendor research, or short directional tests. Reporting tools are early. OpenAI's Ads Manager is in beta, Google does not break out AI Overview placements cleanly, and attribution across the discovery, ad, and Instant Checkout flows is an unsolved problem for most businesses.

The trust posture is also still being tested in public. OpenAI CEO Sam Altman has stated that the company "will not accept money to influence the answer" ChatGPT gives, and OpenAI's January 2026 announcement repeated that "ChatGPT's answers remain independent and unbiased, conversations stay private, and people keep meaningful control over their experience." In late 2024, Altman publicly described ads as close to a last resort for OpenAI's business model. The move from that position to a live ad pilot is significant, and how users respond will determine how the ad surface develops.

Anthropic's stance is different: "We want Claude to act unambiguously in our users' interests. So we've made a choice: Claude will remain ad-free. Our users won't see sponsored links adjacent to their conversations with Claude." Perplexity walked back its own ad experiments. None of these positions are final. They are bets on what users will accept over time. The platforms that keep ads clearly separated, contextually relevant, and free of pressure on the underlying answer are the ones most likely to keep their users with them as the format matures.

The honest summary

We know enough to act. We do not know enough to relax.

Volume, formats, pricing, and measurement of paid ads in AI will keep changing for the next several quarters. The fundamentals will not. Brands that AI systems can find, trust, and quote will see better outcomes on both organic and paid AI surfaces. Brands that are difficult for AI to retrieve will pay more for less, and may end up watching competitors and affiliate publishers occupy placements that should have been theirs.

The most resilient strategy in 2026 is the boring one. Test paid AI carefully in the channels and markets that are open today. Strengthen the AEO foundation behind every future AI surface. Run them as one program. The compounding advantage goes to the businesses that do the work now, before the placements get more crowded and more expensive.

Frequently asked questions

Are paid ads in ChatGPT live right now?

Yes. OpenAI began testing ads on February 9, 2026 for adult Free and Go users in the United States, and has since expanded the pilot to Canada, Australia, and New Zealand. Plus, Pro, Business, Enterprise, and Education tiers stay ad-free.

How much do ChatGPT ads cost?

Initial pricing was approximately $60 CPM at launch in February 2026. A cost-per-click model was added in April 2026, with advertiser bids reported in the $3 to $5 range. Pricing continues to change as the pilot scales.

Does Google show ads inside AI Overviews?

Yes. Google says ads are eligible to appear above, below, and inside AI Overviews in 200-plus markets where AI Overviews are available, and is testing ads in AI Mode as well. Advertisers cannot target AI Overviews specifically, and cannot opt out of them.

Will Google Gemini have ads?

As of late January 2026, Google DeepMind CEO Demis Hassabis said there were no plans for ads in Gemini. That said, Google's existing ads infrastructure and revenue model make ads in Gemini a likely 2026 development, and several industry trackers have reported planned rollouts.

Does Claude have ads?

No. Anthropic stated in February 2026 that Claude will remain ad-free, and reinforced that position with a Super Bowl campaign positioning Claude as the ad-free alternative to ChatGPT.

Why does AEO matter if AI surfaces also have paid ads?

Because the AI systems serving those ads are also choosing which brands to cite in the answer itself, which products to surface in conversational shopping, and which sources to trust. Seer Interactive's published research shows brands cited in AI Overviews earn 35% more organic clicks and 91% more paid clicks than uncited brands. AEO is what gets a brand into the cited set in the first place, which makes both organic visibility and paid performance stronger.

Can I buy products directly inside ChatGPT?

Yes. OpenAI launched Instant Checkout with Stripe in September 2025, powered by the open Agentic Commerce Protocol. U.S. users can buy from Etsy sellers and a growing list of Shopify merchants directly in the chat. Product ranking inside that flow is separate from advertising.

‍