How to Measure ChatGPT Ads Performance: Ads, Organic Visibility, and Conversion
Measure ChatGPT Ads performance across three layers: ad metrics, business results, and organic AI visibility. Includes UTM setup, a worked ROAS/break-even example, and a weekly report format.
Prefer Maya AI in Google
Highlight our stories in Search, AI Mode & AI Overviews.
Last checked: September 21, 2026
To measure ChatGPT Ads performance, you must answer three separate questions: How often is your ad shown and clicked? What business results do those visits produce? And beyond paid ads, how often is your brand recommended in AI answers?
Read together, these three measurements support budget, content and landing-page decisions. But they don't substitute for one another. An ad impression is not an organic recommendation; and a sale attributed to an ad doesn't, by itself, mean the ad created incremental sales.
This guide builds an actionable reporting layout based on OpenAI's ad documentation and Maya's measurement approach. The numeric campaign example is fictional; it's not a customer result or a performance promise.
Which metrics can you track in ChatGPT Ads?
Per OpenAI's documentation, Ads Manager reporting includes impressions, clicks, spend, CTR, average CPC, average CPM and conversions. Conversion measurement requires additional setup. By adding static UTM parameters to landing-page links, ad traffic can also be tracked in your existing analytics tools. OpenAI: how ChatGPT Ads work.
Build your report in three layers:
| Layer | Question it answers | What to track |
|---|---|---|
| Ad performance | What does paid distribution produce? | Impressions, clicks, spend, CTR, CPC, CPM |
| Business result | What do incoming visitors do? | Purchase, qualified lead, revenue, cost per conversion |
| Organic AI visibility | Are we among the options beyond ads? | Recommendation, mention and citation rates on a fixed question set |
The first layer explains the campaign, the second its economic result, and the third the brand's position in the AI answers you track.
Advertising is not buying an organic AI recommendation
OpenAI states that ads are separate from and clearly labeled apart from ChatGPT answers, and that ads don't influence answers. So you shouldn't add ad impressions to your organic visibility score. OpenAI's approach to advertising. For the academic basis that organic visibility can be measurably improved, see GEO: Generative Engine Optimization (arXiv:2311.09735).
In the same week, your ad clicks and your organic recommendation rate may both rise. This simultaneous movement doesn't prove one caused the other.
For example, that week you may have updated your product page, published a new comparison article, or a competitor's visibility may have changed. In your report, record important content and site changes alongside the campaign start.
For organic measurement, fix the question set, the platforms, the country and the language. When you add new questions, flag it as a scope change.
Decide which customer need to test before the campaign
Launching a campaign with the goal "let's show up in ChatGPT" makes results hard to interpret. Set a more specific hypothesis:
For users looking for a lightweight bag with a laptop compartment, clearly presenting the product's weight, compartment size and delivery info may make the purchase decision easier for the right visitors.
Break this hypothesis into three parts:
| Part | Example |
|---|---|
| Customer need | A lightweight, laptop-protecting bag for the daily commute |
| Ad message | Weight, compatible laptop size, and use case |
| Landing page | Product details, price and delivery terms that confirm the same info |
Tracked questions and competitor answers in Maya can be used to research these needs. But tracking prompts are not a record of actual ad impressions or user conversations.
In OpenAI's ad system, context hints help matching; they are not exact-match keywords and don't guarantee display in specific conversations. So don't treat a question you found in research as a directly purchasable targeting unit. OpenAI: ad matching.
How do sector indexes help the ad plan?
Sector data can show which category and competitive areas you should examine more closely.
In Maya's Cosmetics Sector Index for July 13–19, 2026, La Roche-Posay leads the overall ranking, while Maybelline is first in the makeup group at 23.4%. This example shows that overall sector position alone doesn't explain category-level competition. Maya Cosmetics Sector Index.
The takeaway for the ad plan is to evaluate product categories and customer needs separately. The index helps you decide where to research in more detail. But the rates in the index are not ad demand, CPC, conversion rate or expected ROAS. Learning those requires campaign data.
Build a measurement setup that separates paid and organic visits
Use a consistent tagging standard on ad links. For example:
| Parameter | Example value |
|---|---|
utm_source | chatgpt |
utm_medium | paid_ai |
utm_campaign | tr_backpack_2026_09 |
utm_content | lightweight_message_a |
This is example naming, not a required OpenAI standard. If your analytics tool needs a custom channel mapping, define it separately. To separate paid and organic correctly in GA4, check your channel-group definitions: GA4 Default channel group.
Keep campaign IDs, the landing page and the message variant on record too. Don't put personal information in URL parameters.
Don't automatically count an untagged ChatGPT referral as organic. Tags may have been lost or the link shared. Use a separate category for visits whose source can't be verified.
Also don't expect ad clicks and site sessions to match one-to-one. Measurement definitions, consent preferences, pages not loading, and reporting times can create differences. For large gaps, inspect the measurement setup first.
Define the conversion before the campaign starts
In e-commerce, the main result may be a purchase. In a B2B product, a sales-accepted qualified lead may be more meaningful than a form fill.
For each campaign, define:
- The primary conversion event.
- The system where you'll verify the conversion happened.
- The attribution window after ad interaction.
- How revenue handles tax, discounts and returns.
- The method that prevents counting the same transaction more than once.
When the platform report and the order system show different results, first make sure the same date range and definitions are used. Even one report using the ad-interaction date and another the order date can change weekly totals.
Example: why a 2.0 ROAS isn't success on its own
The following campaign is entirely fictional:
| Metric | Value |
|---|---|
| Ad impressions | 60,000 |
| Clicks | 600 |
| Ad spend | €900 |
| Ad-attributed orders | 18 |
| Revenue from those orders | €1,800 |
| CTR | 1% |
| Average CPC | €1.50 |
| Orders per click | 3% |
| Ad cost per order | €50 |
| ROAS | 2.0 |
Calculations:
- CTR: 600 ÷ 60,000.
- CPC: €900 ÷ 600.
- Ad cost per order: €900 ÷ 18.
- ROAS: €1,800 ÷ €900.
Say the average order value is €100. After product cost, payment, operations and expected returns, assume the pre-ad contribution margin is 40%.
The contribution available per order for ads is €40, while the campaign spends €50. Under these assumptions:
€1,800 × 40% − €900 = −€180
So a 2.0 ROAS does not produce a positive result on first-order economics. Under the same assumptions, break-even ROAS is 1 ÷ 0.40 = 2.5.
Repeat purchases can change the economics; but instead of assuming this, examine it with actual cohort data. Also, the figure computed here is cost per order. For customer acquisition cost, the denominator must contain only new customers.
How do you move from metrics to the next action?
Before diagnosing a change, check the data volume and whether measurement is complete. Then use this table as a starting point:
| Observation | Possible explanation to check | Next step |
|---|---|---|
| Impressions, few clicks | The message or offer may not be relevant enough | Test a clearly different message for the same need |
| Clicks, few sessions | Measurement, redirect or page-load issue | Check the link and analytics setup |
| Visits, few purchases | Ad promise may not match page, price or delivery terms | Review the landing page and checkout flow |
| Orders, negative contribution | Acquisition cost may exceed product economics | Evaluate budget, bid and product mix |
| Good ad results, low organic recommendation | Paid reach and organic presence differ | Run a separate organic source and content review |
These are possible explanations. For example, instead of labeling low conversion "the landing page is bad," also check the product's price, stock status and traffic quality.
In a test, focus on a single main question as much as possible. If you change the message, page, offer and budget at once, the result becomes hard to interpret.
Where does Maya fit in this flow?
Maya's GEO/AEO (organic AI visibility) approach focuses on evaluating connected AI ad activity and conversion events together with the context of organic AI visibility. Prompt tracking, competitor comparison, source analysis and traffic analysis provide different evidence for this evaluation. Available ad data depends on the scope of connected accounts and integrations. Maya platform.
A practical working order might be:
- Review your organic standing on priority customer questions.
- Choose a specific need and message to test in the campaign.
- Match ad spend with conversion results.
- Track organic-visibility changes in the same period separately.
- Record a single decision for next week: continue, change, stop, or collect more data.
This approach lets you discuss ad budget and content budget in the same meeting — without mixing the two channels' results.
What should a weekly report look like?
The report's first screen should answer these five questions:
| Section | Content |
|---|---|
| Goal | Which customer need and which conversion is tracked? |
| Ad result | Spend, clicks, conversions and cost |
| Economic result | Attributed revenue and contribution calculation |
| Organic context | Recommendation and citation change on the fixed question set |
| Decision | What will be done, by whom, when re-evaluated? |
The date range, currency and conversion definition should be visible on the report. The week's most important output isn't just a percentage change, but a decision with clear evidence and an owner.
Start your first campaign with a measurement plan
Choose a priority category, a clear customer need and a defined conversion. Evaluate ad results against your own product economics; track organic AI visibility separately.
Assess your scope of work with Maya. You can start the first conversation by bringing your target category, the customer questions you want to track, and the business result you expect from the campaign.
Related guides
Founder of Maya. Writes about where brand discovery is heading as AI assistants replace the search box.
