2026-09-17, updated 2026-11-11, Advertising · ChatGPT

ChatGPT ads: how they work and how to track them

ChatGPT ads are sponsored placements shown below ChatGPT’s answers to users on the Free and Go plans, sold by OpenAI to advertisers and, according to OpenAI, kept separate from the answer itself. They started as a US test on February 9, 2026. By September 2026 OpenAI was selling them in more than 40 countries, offered a self-serve Ads Manager with CPM and CPC buying, and had begun testing conversational “Sponsored Agents”.

The facts below come from OpenAI’s own pages, checked 2026-09-17. The second half of the post covers tracking: the ads[] fields AnswerLine returns for ChatGPT, a method for advertiser share of voice, and ads on Google’s surfaces. One limit first: on the ChatGPT answers served today ads[] comes back empty, so the method describes the contract, not data you can collect now.

Timeline of OpenAI’s announcements

Markets, formats and buying options changed roughly monthly in 2026, so re-check the sources before relying on any single detail.

Who sees ads

Per OpenAI’s help article Ads in ChatGPT (checked 2026-09-17):

• Ads may appear for users on the Free and Go plans. Plus, Pro, Business, Enterprise and Edu accounts don’t have ads.
• Ads aren’t shown in accounts where the user says, or OpenAI predicts, they are under 18.
• For logged-out users, OpenAI says it “designed the experience to show ads that are appropriate for all ages”. The original February announcement described the test as for logged-in adults, so this is a later detail.
• Users can avoid ads by upgrading to Plus or Pro, or by choosing a free no-ads option with additional limits (fewer daily messages and no access to tools like images). Availability “may vary by region and over time”.

Where ads appear, and where they don’t

From the same help article:

• Ads “can appear below the end of a response”, labeled as sponsored and visually separated from the answer.
• There may be “one or more ad units below a ChatGPT response when there’s a relevant match”, and a unit may feature one or more items from one advertiser or from several.
• Ads are not eligible near sensitive or regulated topics, “including personal health, mental health, or politics”. Political advertising is not currently allowed.
• Advertisers in some sensitive verticals, such as health and financial services, “may be eligible to advertise if they meet our strict eligibility criteria”, but ads still aren’t placed near sensitive user content.
• Temporary Chats show no ads, and during the test ads don’t appear in the ChatGPT Atlas browser.

How an ad is chosen

OpenAI’s help center describes the selection inputs: “the context and intent of your current conversation, the ad’s landing page, title, copy, advertiser-provided context hints and targeting selections, and, when ads personalization is enabled, select signals from your broader ChatGPT experience.” When several ads are eligible, OpenAI considers “a combination of factors, including relevance and advertiser bids” to decide which to show first.

With personalization on, those broader signals can include past chats, memory and how the user interacted with ads. With it off, users “will still see ads based on the context of your current chat thread”. OpenAI’s August 31 post adds that the use of broader context depends “on the country and a user’s settings”.

Landing pages matter in a second way. OpenAI’s crawler documentation lists OAI-AdsBot, which visits pages submitted as ads to check policy compliance and “may also use content from the landing page to determine when it’s most relevant to show the ad to users”. OpenAI says OAI-AdsBot only visits submitted ad pages and its data isn’t used to train foundation models.

Two consequences for measurement. Because matching uses conversation and account context, one answer shows a single serving decision, not what every user sees, so measure frequencies over many runs. And an empty ad slot on a health prompt is expected. It says nothing about competitor spend.

OpenAI’s stated limits on ads

These principles appear in the January, February and August posts and the help center:

• Answer independence: “Ads do not influence ChatGPT’s answers. Ads run on separate systems from our chat model, and advertisers have no ability to shape, rank, or alter ChatGPT’s responses.”
• Conversation privacy: advertisers don’t receive chats, chat history, memories, name, email, precise location or IP address, and get “aggregated, non-identifying information about how their ads perform, such as total views or clicks”.
• User control: users can turn off ad personalization, hide or report ads, see why an ad was shown, and clear ads data, which OpenAI says is retained for up to 30 days before removal.

These are OpenAI’s descriptions of its own system and can’t be verified from outside. The separation claim does have a consequence brands can check in their own data: if the systems are independent, a brand’s organic citation rate shouldn’t systematically rise or fall with its ad presence over time.

How advertisers buy

From OpenAI’s May 5, August 31 and September 16 posts, and the help article Create Campaigns for ChatGPT Ads, all checked 2026-09-17:

OpenAI’s advertiser FAQ also states that ChatGPT Ads, as an early platform, doesn’t yet have performance benchmarks across advertisers, industries or campaign types.

What this means for brands

1. Paid and organic results now appear on the same screen from separate systems. On a commercial prompt, a buyer may see your competitor’s ad under an answer that cites you, or the reverse. OpenAI says the two are independent, so buying ads isn’t a documented route to being cited, and being cited doesn’t earn ad placement. Plan and report them as separate channels.

2. Ad exposure in your category depends on plan mix and market. Only Free and Go users see ads, only in markets where OpenAI sells them, and never on sensitive topics. A health brand’s prompts may carry no ads at all. Check availability before building a forecast.

3. Selection starts from the conversation’s intent. Prompts that express a purchase decision (“which X should a Y buy”) are where ads are designed to show, per OpenAI’s framing of “active and decision-oriented” conversations. Build your monitoring prompt set from those.

4. Landing pages affect relevance. OAI-AdsBot reads submitted landing pages, and OpenAI lists the landing page among selection signals, so a thin landing page can hurt delivery as well as conversion.

5. OpenAI reports on your own campaigns only. To see who else appears on your prompts, how often and with which messages, you need to collect answers yourself.

6. Formats, buying options and markets changed monthly in 2026. Date every observation.

Tracking ChatGPT ads

Definitions

• Served ad: an ad included in ChatGPT’s ad response for an answer, whether or not it was displayed.
• Rendered ad: the served ad displayed as a visible card.
• Ad presence rate: the share of answers in a prompt set in which a given advertiser is served (or rendered, reported separately).
• Paid share of voice: an advertiser’s rendered ads divided by all rendered ads across the prompt set and period.

The request

Add include.ads to a ChatGPT request:

curl -X POST https://api.answerline.dev/v1/monitor/chatgpt \
  -H "Authorization: Bearer $ANSWERLINE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "prompt": "best project management tools for small agencies", "country": "US", "include": { "ads": true } }'

For prompt sets, send the same payload as CHATGPT tasks in a batch (shown below).

The ads[] fields

When include.ads is true, result.ads is an array with one entry per served ad. On the ChatGPT answers served today it comes back empty. The shape, per the API reference:

{
  "brand": { "name": "…", "url": "…", "favicon": "…" },
  "cards": [ { "title": "…", "body": "…", "url": "…", "image": "…" } ],
  "visibility": "allowed",
  "adsResponseIndex": 0,
  "rendered": true
}

Use rendered for exposure. Served-but-not-rendered entries show which advertisers were in contention for a prompt even when another card was displayed. Label them “served”, not “seen”.

A tracking run

1. Build the prompt set from your commercial prompts: category (“best X for Y”), comparison (“X vs Y”) and purchase intent (“where to buy X”). Leave out sensitive prompts. Designing a prompt set covers the rest.
2. Choose the markets where you advertise or plan to. ChatGPT answers from 213 countries, including pilot markets such as the US, Canada, Australia, the United Kingdom, Brazil and Japan.
3. Submit the batch and collect results. For recurring runs, put a webhook on each task instead of polling (webhooks).

import collections
import datetime as dt
import os
import time
import requests

API = "https://api.answerline.dev"
HEADERS = {"Authorization": f"Bearer {os.environ['ANSWERLINE_API_KEY']}"}
PROMPTS = {
    "pm-small-agencies": "best project management tools for small agencies",
    "pm-vs": "Asana vs Monday for a marketing team",
    "time-tracking": "which time tracking app should a consultancy buy",
}
MARKETS = ["US", "CA", "AU"]
day = dt.date.today().isoformat()

tasks = [
    {
        "taskType": "CHATGPT",
        "payload": {"prompt": p, "country": c, "include": {"ads": True}},
        "idempotencyKey": f"ads-{pid}-{c.lower()}-{day}",
    }
    for pid, p in PROMPTS.items()
    for c in MARKETS
]
batch = requests.post(f"{API}/v1/async/task/batch", json=tasks, headers=HEADERS, timeout=60)
batch.raise_for_status()
ids = [r["task"]["id"] for r in batch.json()["results"] if r["success"]]

results = {}
while len(results) < len(ids):
    for tid in ids:
        if tid in results:
            continue
        data = requests.get(f"{API}/v1/async/task/{tid}", headers=HEADERS, timeout=30).json()
        if data["task"]["status"] in ("COMPLETED", "FAILED"):
            results[tid] = data
    time.sleep(10)

4. Aggregate served and rendered ads per advertiser:

served = collections.Counter()
rendered = collections.Counter()
creatives = collections.defaultdict(set)
answers = 0

for data in results.values():
    if data["task"]["status"] != "COMPLETED":
        continue
    answers += 1
    for ad in data["response"]["result"].get("ads", []):
        name = ad["brand"]["name"]
        served[name] += 1
        if ad["rendered"]:
            rendered[name] += 1
        for card in ad["cards"]:
            creatives[name].add(card["title"])

total_rendered = sum(rendered.values())
for name, count in served.most_common():
    print(
        f"{name:30} served {count / answers:6.1%} "
        f"rendered {rendered[name] / answers:6.1%} "
        f"paid SoV {rendered[name] / total_rendered if total_rendered else 0:6.1%} "
        f"creatives {len(creatives[name])}"
    )

Store the raw task JSON with the date, prompt id and market. Card titles and bodies over time are a record of competitors’ messaging tests.

A worked example, with invented numbers: 30 prompts in one market give 30 answers. Brand A is served in 12 and rendered in 7; brand B is served in 9 and rendered in 5. Brand A’s served presence rate is 12 / 30 = 40%, its rendered presence rate 7 / 30 = 23.3%, and its paid share of voice 7 / (7 + 5) = 58.3%.

What to do with the data

• Paid vs earned: compare ads[] with organic sources[] and brand mentions in text from the same answer. A competitor rendered in ads but absent from citations is buying coverage it has not earned organically.
• Creative tracking: cards[].title and cards[].body per brand over time show which messages advertisers test.
• Market entry: run your category prompts in a pilot market before launching there.
• Filtered ads: report visibility: "hidden" entries apart from allowed ones.

For the organic side of the same answers, see how AI engines choose citations, measuring brand mentions in ChatGPT, AI share of voice and ChatGPT shopping cards.

Ads on Google surfaces

Google’s surfaces carry ads too, and they come back filled today, each with its own shape.

AI Overview ads need include.aioverview on the Google request, which adds $0.00071 to the $0.0015 Google Search request when an overview is shown. See Google ads, tracking Google AI Overviews, SERP feature tracking and AI Mode shopping results.

Cost

A ChatGPT request is $0.0015 at the Starter tier, and include.ads and include.shopping are free. So a task with ads, or with ads and shopping, is $0.0015.

New accounts get 100 free requests on any engine. See pricing for tier prices and cost planning for budgeting.

Pitfalls

• Treating served as seen. Only rendered: true was displayed. Report served and rendered separately.
• Reading adsResponseIndex as screen position. It is served order.
• One run per prompt. Ad selection depends on context, so single runs swing. Aggregate across runs and days.
• Mixing field shapes. ChatGPT ads[] has brand and cards; Google ads[] has title, url and domain. Normalize to advertiser domain before combining.
• Sensitive prompts. OpenAI says ads are not eligible there, so leave them out of paid share-of-voice denominators.
• Assuming a fixed market list. OpenAI has changed pilot markets over time.

Field definitions are in the API reference and on the ChatGPT engine page.

Keep reading