ChatGPT ads and organic AI recommendations can appear in the same conversation, but they do completely different jobs. A paid placement can reach a buyer during a relevant moment; an earned mention means the model independently assessed your brand as a credible answer. You cannot buy one to substitute for the other, and measuring them requires entirely different approaches.
That was the central argument of an on-demand webinar published by Search Engine Journal, featuring Go Fish Digital and OpenAI. It covered how ChatGPT ads work, how to build earned AI visibility through GEO (generative engine optimization, meaning the practice of helping AI models recommend your brand in their answers), and why standard last-click attribution misses most of the picture.
Are ChatGPT Ads the Same as Organic AI Recommendations?
No. Abhilash Edathil, from OpenAI’s monetization team, demonstrated a travel-planning conversation in which a labeled ad appeared alongside, but visually separate from, the organic answer. He confirmed that ads can draw on the context of the conversation and, where users permit it, personalization data. He was explicit: ads do not inform the model’s answer.
AJ of Go Fish Digital framed the distinction clearly. An organic recommendation is earned visibility; a labeled ad is paid visibility. Two different brands can occupy both positions in the same response without contradiction. The model is not for sale. The ad slot is.
According to Edathil, as of the webinar, ChatGPT ads are shown to eligible adults using the Free and Go versions of ChatGPT, not paid subscribers. Availability by market and vertical was described as still evolving, and he directed advertisers to OpenAI’s ads manager to check current eligibility before testing.
How Do You Measure GEO When AI Answers Keep Changing?
A single AI response is too variable to treat like a traditional search ranking. AJ recommended a three-part measurement framework: Presence (is the brand mentioned for a relevant buyer question?), Representation (is that description accurate and current?), and Competitiveness (how often does the brand appear relative to relevant rivals?).
To make this operational, he suggested starting with roughly 20 to 40 questions that buyers actually ask, running them repeatedly over one to two weeks, and keeping the model and settings consistent across runs. The goal is to identify recurring gaps, not to treat any single answer as a verdict.
Patrick Algrim of Go Fish Digital added a quality dimension: mention count matters less than whether the recommendation is substantive. If the evidence available about a business is thin, the model can only produce a thin explanation. The quality of what an AI system can say is shaped by the quality of the evidence it can access.
Why Fix Your Website Before Chasing Off-Site Mentions?
When asked whether to prioritize on-site content or external mentions, both Go Fish Digital speakers started with the site. Algrim used a practical example: a moving company that offers cross-country moves but never states that on its website should not expect an AI system to infer the service. The pages need to explicitly answer the questions a buyer would ask, describe what the business actually does, and stay consistent with whatever the business says elsewhere.
As Algrim put it at the 33:44 mark of the webinar: “Start with your website because you can control it. Make sure everything is factually true, connected to a cohesive story about your brand, product, service, what it is that you offer.”
AJ added a technical first check: confirm that relevant pages are accessible to search crawlers (the automated programs that fetch and index web content) and not blocked by robots.txt rules (a file on a website that tells crawlers which pages to skip) or a CDN (content delivery network) configuration. Reviews, PR coverage, and other external signals still matter, but they cannot repair an unclear or contradictory on-site account of the business.
How Should You Read AI Referral Data in Analytics?
AI influence often disappears from standard last-click reporting. A buyer may discover a brand through a chatbot, then search for it separately, or paste a URL directly into a browser. GA4’s AI-assistant referral data captures only part of that path.
AJ recommended supplementing referral data with customer self-reporting and trends in branded and direct traffic, while keeping AI-assistant sessions separate in reports. Algrim’s suggested method: test specific website or campaign changes over a defined period and track what shifts, rather than assigning every increase to a single source without evidence.
For ChatGPT ads specifically, Edathil recommended deciding on the desired outcome first (reach, traffic, or conversions), then connecting conversion data through OpenAI’s available pixel or API (application programming interface, the technical connection that passes data between systems), and adjusting creative, bids, or budgets based on results.
Quick-Start Checklist Before Running a GEO or ChatGPT Ads Test
- Write down 20 to 40 questions your buyers actually ask, then check current AI responses for presence, accuracy, and how competitors appear.
- Run those prompts consistently over one to two weeks using the same model and settings before drawing conclusions.
- Audit the pages that should answer those questions: clarify services, eligibility, and what makes you different from competitors.
- Verify those pages are not blocked from being crawled by robots.txt rules or CDN configuration.
- Check that on-site descriptions are consistent with what appears in external reviews, directories, and press coverage.
- If testing ChatGPT ads, confirm eligibility in OpenAI’s ads manager and define a specific outcome before launching.
- Supplement GA4 AI referral data with branded search trends, direct traffic, and customer-reported discovery paths.
What This Means for AI-Search Visibility
Here is the clearest way to think about this: paying for a ChatGPT ad slot puts your offer in front of a buyer during a relevant conversation, but it does not change what the model thinks about your brand. Earning a mention in the model’s answer takes longer and requires consistent, accessible evidence, but it compounds over time regardless of ad spend.
The practical implication is that most businesses are not yet ready to earn consistent AI mentions because the content that would support those mentions is either missing from their site, blocked from crawlers, or contradicted by what the business says elsewhere. Edathil’s point about ads not informing the answer is significant here. Brands that assume a paid ChatGPT campaign will lift their organic AI presence are likely to find they have paid for impressions without building anything durable.
In Hingewise’s assessment, the measurement framework described by Go Fish Digital (Presence, Representation, Competitiveness) is more useful than most GEO scorecards currently in circulation, precisely because it separates three things that often get conflated: whether a brand shows up at all, whether what the model says is accurate, and whether the brand holds its ground against specific competitors. A business could score well on the first and fail entirely on the other two.
One point the webinar did not address, and worth flagging: the weighting of these three dimensions likely varies by category and buyer intent. For a high-consideration purchase such as legal services or enterprise software, the quality of representation may matter more than raw mention frequency. For local services with high competition, competitive frequency may be the more decisive signal. There is no universal formula yet, which makes sustained measurement over time the only reliable baseline right now.
What to watch: how OpenAI expands ChatGPT ads eligibility across markets and verticals, and whether brands begin reporting measurable lift in AI-source referrals after campaigns versus only downstream signals like branded search.
Source: Search Engine Journal, on-demand webinar with Go Fish Digital and OpenAI.
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.