On May 5, 2026, OpenAI launched ads.openai.com. Sponsored cards now appear at the bottom of ChatGPT answers when a question looks like shopping. They are labeled. They cannot change the AI answer above them.
This matters more than a typical product update.
OpenAI has formally split AI-assisted discovery into two layers: what AI recommends, and what brands pay to place next to it. Their own documentation states the relationship plainly: "Ad placement and organic citation are different surfaces, but they share signals."
That sentence will matter to every brand making AI visibility decisions in 2026.
Disclosure: Envoyra sells the AI visibility scanning tool used to collect the data cited in this article. We have a financial interest in findings that support measuring unpaid AI recommendations. The method is set out in full below.
Paid and unpaid, in the same answer
When a buyer opens ChatGPT and asks "what are the best AI meeting assistants for a remote team," two things can happen to a brand in that moment:
1. It appears in the organic recommendation: the body of the AI's answer. Nobody pays for this. It is earned. 2. It appears in a sponsored card below the answer: a paid placement, shown when the question looks like shopping and priced by bidding. Labeled "Sponsored."
The buyer sees both, in the same interface. But they work differently, cost differently and mean different things.
The organic recommendation is what buyers read as the answer. The sponsored placement is what buyers see as the advertisement.
A brand can run ChatGPT ads and still be missing from the unpaid answer above its own ad. Its competitor, the one appearing in the body of the answer, is being recommended. The brand with the sponsored card is paying to appear beneath someone else's recommendation.
That is the scenario brands need to understand before they buy AI advertising.
What the organic layer actually looks like
Envoyra asked Perplexity, Claude and ChatGPT Search a set of buyer questions in AI Meeting Assistants, AI Customer Support Agents and SMB CRM. We collected this data before sponsored placements existed. That makes it a clean baseline of unpaid recommendations.
Three findings describe the unpaid layer.
Each AI assistant names different brands.
The same brand does not show up equally across AI assistants. In AI Meeting Assistants, Claude's mentions were concentrated on two brands, Otter and Gong. On Perplexity, the same questions produced a more spread-out list. Fathom was named noticeably more often on Perplexity than on Claude. That is not a rounding error. Each AI assistant gave a different picture of the category.
A brand that shows up on one AI assistant is not necessarily named on another. Each one has to be measured on its own.
Big platforms crowd into categories, and advertising can't fix that.
In AI Meeting Assistants, Microsoft Copilot, Zoom AI Companion and Teams were named as alternatives in nearly every answer, on all three AI assistants. These are not startups with better content. They are platforms already built into tools buyers use, and AI assistants treat them as valid answers to meeting-assistant questions.
In AI Customer Support, Agentforce, ServiceNow and Salesforce were named in most Perplexity answers. Big platform players were taking space from the independent tools in the category.
Sponsored placements do not change this. A brand cannot buy its way out of a category that AI assistants increasingly treat as a platform market. Responding means showing up in the unpaid answer.
A few brands take most of the mentions.
In SMB CRM, HubSpot and Pipedrive together took most brand mentions on every AI assistant we tested. Zoho, Copper, Close CRM and ActiveCampaign split the rest. The pattern held across Perplexity, Claude and ChatGPT. It is not a quirk of one AI assistant.
In categories this concentrated, brands outside the top two are rarely named at all. That is not a marketing problem. It is a measurement problem. Most of these brands do not know how rarely they are named.
Why organic baseline matters when paid AI arrives
Paid AI advertising changes the math for every brand making AI visibility decisions. Here is how:
If you don't know where you stand unpaid, you can't judge what your ads return.
A brand running ChatGPT ads without measuring its unpaid presence can't tell two very different situations apart: - Its ad adds to an unpaid presence it already has (competitor in the answer, brand in the answer and in the ad below) - Its ad stands in for an unpaid presence it lacks (competitor in the answer, brand only in the ad below)
These call for different strategies and return different results. Without measurement, they look the same.
Unpaid recommendations and paid placements share signals, so unpaid presence may affect paid results.
OpenAI's own wording says the two layers are not fully separate. Brands that are cited more often in unpaid answers may see their ads perform differently from brands with no unpaid presence. The unpaid layer is not just a brand metric. It may also feed ad performance.
The organic recommendation does not disappear when ads exist.
The AI answer does not change because a brand bought an ad. The body of the answer, the recommendation, is produced the same way as before. A brand missing from that layer stays missing, whatever appears below it.
The measurement gap that advertising reveals
AI advertising puts a price on AI visibility. Brands that buy ChatGPT ads without first measuring their unpaid presence won't know what the ads are adding.
The brands most exposed assume AI visibility is the same as search visibility, or that paid placement and unpaid recommendation are interchangeable. They are not. Search had this distinction for twenty years: organic rankings and paid ads have always been separate columns on the same page. AI discovery is now following the same pattern, faster.
The unpaid layer can be measured. Differences between AI assistants, big platforms crowding in, and a few brands taking most mentions are all specific and countable. You can set a baseline before any advertising decision.
Brands that measure unpaid presence first will know whether their ad spend adds to visibility they have or stands in for visibility they lack. That difference matters for both returns and strategy.
Method
The data comes from buyer questions in AI Meeting Assistants, AI Customer Support Agents and SMB CRM, with the same number of questions in each category. AI assistants: Perplexity sonar-pro (searches the live web), Claude Sonnet 4 (answers from training), ChatGPT gpt-4o with web search on. Data collected May 2026 using Envoyra's standard question set. OpenAI Ads details come from OpenAI's public documentation and AdExchanger reporting. Differences between AI assistants compare how much of each assistant's mentions went to its top two brands. Platform crowding counts the answers that named at least one big-platform brand we were not tracking.