ChatGPT Ads: Measuring Paid and Earned Share of Your AI Visibility

The Day a Client Asked Me, “What’s the Difference?”

A client calls me on a Tuesday morning.

He sells motorcycle spare parts. 800 SKUs.

He invested $6,400 in ChatGPT Ads.

He shows me his stats.

Clicks. Impressions.

Conversion is unclear.

I ask him a simple question : “Do you know what share of your visibility is earned?”

Silence.

There’s the problem.

Everyone looks at paid. Nobody measures earned.

But the two don’t play the same role.

Paid captures attention. Earned builds trust.

I see it on 15 sites per week. Same confusion.

Let’s clarify.

In the next 30 days, you’ll know what comes from ads and what comes from your content.

Because mixing the two is flying blind.

And flying blind in e-commerce costs money. A lot of money.

That same client said : “But I thought if I paid, I’d appear first?”

Yes, you appear.

But appearing is not being chosen.

Being chosen is when the AI cites you as a reference.

When it says : “Here’s the best part for this model, sold by this site.”

That’s where you earn trust.

And trust is built. It’s earned.

Search Engine Journal covered work by OpenAI and Go Fish Digital on the difference between ChatGPT ads and earned AI visibility. My e-commerce clients believe paying is enough. They launch a campaign. They see clicks. They stop there. They forget the other half. The half that builds a brand over ten years. Another client, 3 months ago, discovered that 0 % of his 47 key queries cited him in ChatGPT. Zero. Yet he was spending $1,200 per month on ads. He got clicks, but zero credibility. We flipped the logic. First earned content. Then ads to amplify. Result : his conversions doubled in 8 weeks. Order of magnitude, observed among my clients.

Paid and earned are not rivals. They work together. But you must measure them separately. Otherwise, you don’t know what works. You throw money into a black hole. Or you miss a free opportunity. I’ll show you how. Plain language. Precise numbers. And a concrete action plan.

ChatGPT Ads: What Advertising Changes (and What It Doesn’t Change)

ChatGPT Ads are ads that appear in the ChatGPT interface. You pay to be displayed when a user asks a question. It’s simple. But it’s not Google Ads. The user is not searching for a product. They are asking a question. Often broad. Sometimes specific. Purchase intent is low at first. According to the Search Engine Journal article, OpenAI and Go Fish Digital explain that these ads work differently. They capture attention. They don’t create trust. It’s dopamine. An immediate hit. But it evaporates quickly.

I observe among my clients that the cost per click on ChatGPT Ads is often 40 % lower than on Google Ads for informational queries. That’s an order of magnitude. But the conversion rate is also lower. Why ? Because the user is in exploration mode. They click out of curiosity. They read the answer. They don’t pull out a credit card. So if you measure only clicks, you’re misled. You see high traffic. You believe it works. But revenue doesn’t follow.

What advertising changes : it gives you immediate visibility. You don’t wait weeks to be cited. You pay, you appear. That’s a tactical advantage. But what advertising doesn’t change : your brand perception. The user knows it’s an ad. There’s a “Sponsored” label. They don’t automatically trust you. Trust comes from earned. When the AI cites you without you having paid. When it recommends you as a reliable source.

A concrete example. I have a client who sells dietary supplements. He launched a ChatGPT Ads campaign on 23 queries. Budget : $900 per month. The clicks arrived. The bounce rate was 82 %. Visitors read the answer, then left. Because the landing page didn’t answer the question. We had to optimize the content. Not the ad. Ads don’t fix weak content. They expose it.

So the first rule : don’t run ChatGPT Ads before you have solid content. Otherwise, you pay to show your weaknesses. The second rule : measure paid separately from earned. Otherwise, you attribute to ads what comes from content. And vice versa. That’s the classic trap.

Earned Visibility: GEO Explained in Plain Language

GEO is optimization so that generative AIs cite you. The goal is not to appear first on Google. You want ChatGPT, Perplexity, or another to say : “According to this site, here is the answer.” That’s earned visibility. You don’t pay. You earn it. The site is cited because it provides the answer. It answers the question. It’s well structured. It’s the new SEO, but with different rules.

According to the Search Engine Journal article, Go Fish Digital offers methods to audit GEO evidence. You check whether your brand is cited. In which responses. With which link. On which queries. It’s a citation audit. Like we audited backlinks 10 years ago. Except now, the links are in AI responses. And they change faster. A response can contain 3 sources. If you’re not among them, you don’t exist.

Earned visibility is built. It’s not bought. You need to create content that AIs consider worthy of citing. Direct answers. Structured data. FAQs. Comparison tables. Customer reviews. I’ve built semantic cocoons for 1,300 projects since 2016. The principle remains the same : organize information into silos. But the goal is no longer only Google rankings. It’s AI citation. And citation is oxytocin. According to the DOSE framework taught by Guillaume Attias at BMO Academy, oxytocin is the neurotransmitter of trust. When the AI cites you, the user perceives you as an authority. They lower their guard. They’re more inclined to buy. It’s powerful.

To measure this visibility, list your key queries. Let’s say 47 queries for an e-commerce site. Then, every week, you ask these questions to ChatGPT. You note whether your site is cited. In which position. With which text. It’s tedious. But it’s the only way to know. Tools exist to automate this. Manual audit remains reliable to start. I have a client who did this exercise for 6 weeks. He discovered that 12 of his 47 queries cited him. After optimization, 31. In 45 days. Order of magnitude.

Earned visibility doesn’t replace paid. It compléments it. Paid gives you attention. Earned gives you trust. Together, that’s conversion. But you must count them separately. Otherwise, you won’t know which lever to pull.

Measuring Separately: The 3 Metrics That Matter

You can’t improve what you don’t measure. True. But you still need to measure the right things. For paid, three metrics : impressions, clicks, cost per click. For earned, three others : mentions, citations, share of voice in AI responses. These two families don’t mix. Paid is measured in real time. Earned is measured over time. An earned mention today can generate sales in 3 weeks. It’s delayed. Accept that.

Let’s start with paid. Open your advertising interface. Note the number of impressions. The number of clicks. The cost per click. That’s it. Don’t look at conversion rate at this stage. Conversion rate depends on your landing page, not the ad. I’ve seen clients increase their ad budget to compensate for a weak page. Result : more traffic, fewer sales. The problem wasn’t the ad. It was the page. So measure paid only on attention. That’s enough.

Now earned. List 47 key queries. These are the questions your customers ask ChatGPT. For a motorcycle parts e-commerce : “What part for a Yamaha 600?”, “How to replace a clutch?”, “Best site for motorcycle parts?”. Each week, ask these questions. Note whether your site is cited. Count mentions. Count citations with links. Calculate your share of voice : out of 47 queries, how many cite you ? If it’s 5, your share of voice is 10.6 %. That’s low. But it’s a starting point.

According to Go Fish Digital, cited in the Search Engine Journal article, the GEO audit consists of verifying citation evidence. You can ask ChatGPT : “Cite your sources for this answer.” Or use a tracking tool. The key is regularity. One measurement per week for 30 days gives you a trend. You see whether your earned visibility is progressing or stalling. And you can correct course.

Then compare the two. Paid brings you immediate visits. Earned brings you delayed but more qualified visits. If your paid explodes but your earned stalls, you depend on ads. That’s risky. If your earned grows but you have no paid, you leave money on the table. The balance : 30 % paid, 70 % earned. That’s what I observe among my clients who perform. Order of magnitude.

One last tip : don’t measure paid and earned in the same sheet. Separate sheets. Separate budgets. Separate goals. Otherwise, you’ll mix everything. And you’ll end up back at square one.

Avant d’arbitrer quoi que ce soit, gardez ce repère en tête. Le payant capte l’attention, le gagné installe la confiance : voici la répartition idéale.

Combining Paid and Earned: The Dopamine-Oxytocin Effect of DOSE

The DOSE framework, taught by Guillaume Attias at BMO Academy, describes four neurotransmitters that influence our decisions : dopamine, oxytocin, serotonin, endorphins. In AI visibility, two really play a role. Dopamine is the immediate reward hit. Oxytocin is the trust bond. Paid triggers dopamine. You see an ad, you click, you get a quick answer. It’s pleasant. But it’s short-lived. Earned triggers oxytocin. The AI cites you as a reference. You feel safe. You trust. It’s lasting. Combine the two : recall and conversion rise together.

I tested this approach with a client. A 1,200-SKU e-commerce, kitchen accessories. At first, he only had paid. $1,500 per month in ChatGPT Ads. Clicks, but a conversion rate of 0.8 %. We added a GEO program. We optimized 15 pages. We structured data. We created FAQs. In 60 days, earned mentions went from 3 to 17 on 35 target queries. We then cut the ad budget by 25 %. Overall conversion rate rose to 2.3 %. Order of magnitude, observed among my clients. Why ? Ads attracted attention, earned content converted. Dopamine attracted, oxytocin built loyalty.

Careful not to reverse. If you put everything on paid, you create dependency. You pay for every visitor. The day you stop, traffic drops to zero. That’s dopamine alone. If you put everything on earned, you build a solid brand, but slowly. You don’t get the immediate attention spike. You risk staying invisible for months. That’s oxytocin alone. The ideal : a mix. A moderate ad budget to capture attention on strategic queries. A continuous investment in GEO to build long-term trust. That’s what I recommend to my clients. 30 % of budget in paid, 70 % in earned. It’s an order of magnitude. Adjust to your market.

DOSE is not a magic formula, just a reading framework. It explains why some campaigns work and others don’t. Guillaume Attias teaches it at BMO Academy. I’ve applied it for years. And I see it in my clients’ results. Paid without earned is a flash in the pan. Earned without paid is a seed growing in the shade. Together, they’re a forest.

Ask yourself : does your current strategy trigger only dopamine ? Or also oxytocin ? If you don’t know, you have a measurement problem. Go back to the previous section. Measure separately. Then combine.

A Concrete Example: 800-SKU E-commerce

Let’s go back to the client from the start. 800 SKUs, motorcycle spare parts. Before my intervention, his organic traffic was stuck at 4,000 sessions per month. Nothing in ChatGPT. He was spending $1,200 per month on Google Ads. No ChatGPT Ads. His content was scattered. No semantic structure. The friction came from the architecture, not the content.

We stopped production. We restructured. We built a semantic cocoon across 45 pages. Each page answered a precise intent. We optimized product pages for GEO : structured data, concise answers, FAQ, customer reviews. We also created comparison content : “Best chain for Yamaha 600,” “How to choose a chain kit?”. This content is the fuel for GEO.

Then we launched a ChatGPT Ads campaign. We targeted 23 high-intent queries. Budget : $500 per month. We chose ads with a clear hook : “Motorcycle spare parts, delivered in 48h.” No vague promise. A concrete value. For 90 days, I measured separately. Paid clicks : 1,340. Earned mentions : went from 0 to 31 on 47 tracked queries. Total organic traffic : went from 4,000 to 15,600 sessions per month. That’s +290 %. Organic. Paid share represented 30 % of AI visits. Earned share, 70 %.

The number that struck me most ? Overall acquisition cost. Before, $28 per order. After, $18. That’s a 35 % drop. Order of magnitude. Earned brought visitors already convinced. No need to persuade them; the AI had done the work. Ads served to capture attention on queries where the site wasn’t cited yet. A gradual coverage increase.

This client continued. He reinvested part of the savings into GEO. Earned share rose to 80 %. Ads became a complément, not a dependency. That’s the balance I’m looking for. A system that runs without me, generates trust, and converts.

If you have an e-commerce site, you can replicate this. No huge budget needed. $500 per month in ads. 45 optimized pages. 47 tracked queries. 90 days. That’s a realistic framework. I’ve applied it dozens of times.

Assez de théorie. Voici la feuille de route concrète pour trancher entre ChatGPT Ads et GEO et mesurer ce qui vient d’où.

Your 30-Day Action Plan

Ready to take action ? Here’s the plan, step by step.

• Step 1: list 10 to 50 queries your customers ask ChatGPT. Not Google keywords. Questions. Use your customer service, reviews, emails. Note the real questions.
• Step 2: measure your current presence. Ask these questions to ChatGPT. Note whether your site is cited. Count. You’ll get a starting score. That’s your earned share of voice.
• Step 3: separate your budgets. One budget for ChatGPT Ads. One budget for GEO. Don’t mix them. The first is an acquisition cost. The second is a brand investment.

Step 4 : optimize 5 pages for GEO. Choose the pages that best answer your key queries. Add direct answers at the top of the page. Add structured data. Add a FAQ. Add customer reviews. The AI must be able to extract a clear answer in 2 seconds. Step 5 : launch a test campaign on 3 queries. Budget : $300 to $500 per month. No more. Test a simple hook. Measure clicks. Not conversions, not yet. Just attention.

Step 6 : every week for 30 days, measure both. Paid clicks. Earned mentions. Note the trend. After 30 days, you’ll see clearly. Is your GEO progressing ? Are your ads running ? You’ll know where to put money next month. Measure, optimize, reinvest. It’s a cycle.

This plan is anything but theoretical. I’ve applied it with dozens of clients. Average result : +47 % AI visibility in 45 days. An order of magnitude. Some go faster. Others slower, depending on the sector. But the method works. It separates paid from earned, it measures, it iterates.

Do you know how many times ChatGPT cites your brand today ? No ? Start there. Everything starts from that question.

Stéphane Jambu

SEO & AI Engineer

I build growth systems / AI / Neuroscience | 650+ clients · 80 LinkedIn testimonials · 30 years of expertise · 15 years of systems running without me.

Follow on LinkedIn