LLM Ad Platforms in 2026: Where to Spend Now, Where to Wait, and How They Compare to Google Ads and Social

Every week another client asks me some version of the same question: should we be advertising in ChatGPT? The honest answer is that “AI ads” is not one channel. It is a handful of very different platforms at very different stages of maturity, and only some of them deserve budget in 2026. After 20+ years in paid media, my rule is simple: put money where you can target, measure, and control, and test carefully everywhere else.

Below is how the major LLM ad environments work today, how they compare to Google Ads and paid social on targeting, reporting, optimization, and setup, and where I would prioritize versus wait. This market is changing fast, so check platform documentation before you commit budget.

1. The Short Answer: Priorities for 2026

• Prioritize now: Google’s AI surfaces (AI Overviews and AI Mode) through the Google Ads account you already run, and Microsoft Advertising, which now includes Copilot and strong B2B targeting.
• Test with a capped budget: ChatGPT Ads, now open to US businesses through a self-serve beta.
• Wait: Amazon Ads inside ChatGPT (limited pilot), Google’s newest Gemini-built formats (still in testing), and Copilot’s checkout and offer formats (no dates given).
• Not an ad platform: Perplexity. It pulled advertising from its search product, so treat it as an organic visibility channel.

2. Platform by Platform

Google AI Overviews and AI Mode (prioritize). Ads here run through your existing Google Ads account, so there is no new platform to learn. Google’s May 20 announcement describes Direct Offers (a pilot that began in January 2026) and new Gemini-built formats, Conversational Discovery ads and Highlighted Answers, that are still in testing. Google recommends building the foundation with AI Max for Search, AI Max for Shopping, and Performance Max, and says advertisers upload promotions and guardrails inside Google Ads. Shopping-style ads with generated explainers and a “Business Agent for Leads” chat agent are listed as coming in the following months. The upside is that you keep Google’s keyword, audience, and conversion infrastructure. The downside is the one I keep writing about: the more of this runs through AI Max and Performance Max, the less you see about where your ads showed.

Microsoft Advertising and Copilot (prioritize, especially for B2B). Microsoft’s Activate 2026 event (May 19), as summarized by ALM Corp, described LinkedIn-informed professional targeting expanding for B2B advertisers, natural-language Audience Generation, and Performance Max reporting that now includes search term insight, landing page reporting, and placement visibility. Copilot-specific formats such as Offer Highlights and Copilot Checkout were discussed without availability dates, and Clarity added AI Visibility reporting that shows how your content appears in AI-generated answers. The strategic point for B2B: Microsoft is the one major platform where professional-profile targeting and AI placement are being built together, and you can usually import campaigns from Google Ads rather than starting over.

ChatGPT Ads (test, do not scale yet). OpenAI launched a beta self-serve Ads Manager for US businesses on May 5, with signup at ads.openai.com, account verification, a credit card, and no agency required. Ads appear below relevant conversations for Free and Go users, not inside the answer. Since launch OpenAI has added CPC bidding alongside CPM, daily and lifetime budgets, geo-targeting by state, DMA, and ZIP code, custom audiences uploaded from email or phone lists, a pixel and Conversions API, and conversion-optimized campaigns. On October 5 it announced new measurement integrations, brand-suitability work, a visual ad format for image generation, and early partner results. Those results are partner-reported, so read them as encouraging, not proven. I would run a small test, but not make it a core channel yet.

Amazon Ads in ChatGPT (wait). Amazon announced on September 10 that select US advertisers can extend Amazon Ads campaigns into ChatGPT as part of a limited pilot. Unless you are in that pilot, there is nothing to buy yet, and the announcement does not describe targeting or reporting. If you sell through Amazon, ask your rep about eligibility; otherwise watch and wait.

Perplexity (not an ad platform). Perplexity tested ads in 2024, then removed advertising from its search product, saying ads undermined trust in its answers. Plan on earning citations with authoritative, well-structured content instead of paying for placement.

3. How LLM Ads Compare to Google Ads and Social

	Google Ads	Paid Social (Meta, LinkedIn)	ChatGPT Ads (today)
Targeting	Keywords, intent, audiences, locations. Increasingly AI-expanded (AI Max, Performance Max).	Audience-first: demographics, interests, job title and company (LinkedIn), lookalikes, retargeting.	Conversation context and intent. No exact-match keywords; advertisers add “context hints” per ad group. Geo down to ZIP code and uploaded custom audiences. Delivery is controlled by OpenAI.
Reporting	Deep, though Performance Max and AI Max limit search-term and placement detail.	Detailed audience, placement, and creative reporting inside each platform.	Impressions, clicks, spend, CTR, CPC, CPM, and conversions (with pixel or Conversions API), via tables, charts, and CSV. Aggregated only; no conversation data.
Optimization	Mature: Smart Bidding, tROAS, tCPA, experiments.	Mature: conversion objectives, creative testing, learning phase.	CPM, CPC, and a conversion objective. “Maximize results” bidding is preselected on new ad groups and OpenAI says it does not guarantee a specific CPA, CPC, or ROAS.
Setup	Established. Complex, with many levers.	Established. Creative-heavy.	Simple: account, verification, billing, favicon, short text ad plus image. Verification can delay launch.

The big conceptual shift is that Google captures stated intent (a search), social creates demand from who a person is, and LLM ads respond to what someone is working through in a conversation. That can be valuable for B2B buyers researching a purchase, but it also means you are steering with hints and creative, not with keywords and bid modifiers.

4. Where I Push Back: Control and Data Visibility

• You cannot see the conversation. That is good for privacy, but it also means you cannot audit the context your ad was matched to. Reporting is aggregated, so you are trusting the platform’s judgment on relevance.
• Automation is the default. ChatGPT’s “Maximize results” bidding is preselected, with budget as the main limit and no guaranteed cost target. A manual max bid is the one documented hard ceiling. Change the default deliberately.
• Reporting is built on partners. Strong attribution and incrementality will depend on third-party tools, which adds cost and setup. That is a reasonable trade, but it should be a conscious one.
• Early numbers are mostly self-reported. Early CPM and click-through-rate figures circulating from the spring reflect a pilot that has since changed its pricing and access. Do not benchmark against them.

5. What This Means for B2B Distributors and Manufacturers

Industrial buyers do research in AI tools: specifying a part, comparing suppliers, checking compatibility. That is a real opportunity, but your sales cycle is long and your conversions are RFQs, not checkouts. ChatGPT’s current formats and its Free and Go audience skew consumer. Google and Microsoft, where you already have keyword and audience control, plus LinkedIn-informed targeting on Microsoft, are where I would put the money first. Use ChatGPT as a learning test, and use Perplexity as an organic-content opportunity.

6. The Playbook

1. Fix measurement first. Make sure qualified RFQs and closed orders are tracked and imported into Google Ads and Microsoft Advertising. See my guide to tracking RFQs from Google Ads.
2. Audit your AI exposure in Google. Review AI Max and Performance Max settings, brand controls, and asset-level reporting before expanding into AI Mode formats.
3. Import into Microsoft Advertising. Bring over your best Google Ads campaigns, then layer on LinkedIn-informed targeting and review placement reporting.
4. Test ChatGPT Ads with a capped budget. Use a single product line, 4 to 8 weeks, a manual max bid, geo limits around your service area, and a pixel or Conversions API in place before launch.
5. Write for context hints. Describe the problems and specs your buyers discuss, not just product names.
6. Send your own customer lists carefully. If you use custom audiences, upload only data you have the right to use and exclude existing customers when prospecting.
7. Build for citation. Publish specs, comparison pages, and FAQs that AI tools can quote. This works across ChatGPT, Gemini, Copilot, and Perplexity at no media cost.
8. Review quarterly. Reassess each platform every quarter against cost per qualified quote and closed revenue.

Sources

• OpenAI on new ways to buy ChatGPT Ads
• OpenAI on its new ad format and measurement tools
• PPC Land on the ChatGPT Ads Manager opening to US businesses, its budget and geo-targeting update, and the Maximize results bidding default
• Google on new Gemini-built ad formats and Direct Offers
• Amazon Ads on its ChatGPT advertising integration
• ALM Corp on Microsoft Advertising Activate 2026
• The Decoder on Perplexity removing advertising

Not sure which AI ad platforms deserve a share of your budget? Let’s look at your numbers and build a test plan that protects your control.