ChatGPT Ads Scraper

What does ChatGPT Ads Scraper do?

ChatGPT Ads Scraper collects the sponsored ads ChatGPT shows under its answers. Use it as your own ChatGPT ad library and ChatGPT ads tracker: spy on ChatGPT ads from any brand, model or topic, check context hints, track competitors' ChatGPT ads, find brands advertising on ChatGPT.

Use cases

• Competitor monitoring. See which brands advertise for your keywords, which products they push, and at what price.
• Market research before buying ads. Find out how crowded a category is before you spend on ChatGPT ads.
• Ad verification. Check that your own ads show up, for which queries, and with the right details.
• Context Hints checking. See which conversations trigger your ads or a competitor's, to check if Context Hints match the queries you want.
• Trend tracking. Schedule daily runs over the same queries and build a history of the ChatGPT ad market.
• Ad creative research. Collect the headlines, ad copy, images, and prices competitors use, to inspire your own ChatGPT ads.
• Agency reporting. Show clients who advertises in their category on ChatGPT, and pitch ChatGPT ads with real examples.

How to use it

1. Open the Input tab.
2. Add queries to Queries: any topic advertisers pay for, like products, travel, software, or finance.
3. Click Start.
4. Open the Output tab when the run finishes.

Input

Field	Description
queries	What to ask ChatGPT. Every ad shown in the chat is saved. Name a brand or model ("best Nike running shoes", "Nike Pegasus 42") to steer the chat toward it.
depth	How many chats each query gets: quick (8 chats, about $0.36 per query), standard (24, about $1.08, default), or deep (60, about $2.70). Plus $0.015 per ad. The run ends when every query has had its chats, so 3 queries at Standard means 72 chats. Deeper finds more ads and advertisers.
country	Two-letter country code, like US, GB, or DE, picked from a list. Chats run from this country, so you see the ads shown there. Default US, where ChatGPT shows the most ads. Write your queries in English: follow-ups are in English too, whatever the country.
saveScreenshots	Saves proof of each ad: a screenshot with the end of ChatGPT's answer above it, and the ad's HTML with all the data ChatGPT sent. Linked from screenshotUrl and htmlUrl. Default off. Adds a few seconds per ad, no extra charge.
followUpQueries	Advanced. Sent in order after each query, the same for every query. Leave empty to use the defaults. Keep them about buying, like "Which one should I buy?". Up to 8.

See which brands advertise for a topic

{
    "queries": ["best trail running shoes"]
}

Find ads from one brand or model

Name the brand or model in your queries. A brand alone, like "monday.com", shows who advertises on its name: the brand itself, competitors, and stores. Ads from all advertisers are saved, so filter the results by advertiser.

{
    "queries": ["monday.com"]
}

Ask your own follow-ups

Write the follow-ups when you want to know exactly which messages bring ads, or when the defaults don't fit, like booking a trip. Every chat sends the same ones, so each ad's turn and message show which follow-up brought it.

{
    "queries": ["cheap flights to Lisbon"],
    "followUpQueries": [
        "Which one would you recommend booking?",
        "Where can I book it at the best price?",
        "Are there any travel deals right now?"
    ]
}

Check that your own ads show up

Use the queries your customers ask about your product and turn on Save ad proof. Filter the results by landingDomain to find your ads, with a screenshot of each.

{
    "queries": ["monday.com"],
    "saveScreenshots": true
}

Compare countries

Run the same queries with a different country, like "country": "GB", and compare with the US run. Expect fewer ads outside the US.

Track ads over time

Save your input as a task and schedule it daily. Count unique ads by adId to see which ads are new and which keep running.

Output example

You get one dataset item per ad:

{
    "query": "best running shoes",
    "turn": 3,
    "message": "Where can I buy it at the best price?",
    "country": "US",
    "chatId": "6ac5220b-ddd0-83ea-ab6b-bb82b700ad5c",
    "adId": "3f9a1c2b7d4e5f60",
    "openAiAdId": "ad_a62b73c023908193ba2dc704b277853d",
    "campaignId": "cmpn_6aa882c6be80819b9a23c010e8c40f3c",
    "adGroupId": "adgrp_69d0c046a988819b97c5f7e7305dec57",
    "impressionId": "06ac521c-3dfc-72cc-8000-944fa08ccf6b:0",
    "position": 1,
    "advertiser": "Nike Inc.",
    "brand": "Nike Inc.",
    "title": "Nike Men's Run Swift 3 Road Running Shoes in Grey, Size: 8",
    "body": "No matter the run, the Swift 3 helps you get out the door...",
    "price": 85,
    "currency": "USD",
    "priceText": "$85",
    "priceSource": "ad",
    "originalPrice": 100,
    "discountPercent": 15,
    "rating": 4.7,
    "imageUrl": "https://files.openai.com/content?id=file_...",
    "url": "https://www.nike.com/t/run-swift-3-mens-road-running-shoes-SBE3UoK3/IB3972-021?...",
    "landingDomain": "nike.com",
    "format": "product_card_v2",
    "displayType": "single_advertiser_ad_unit",
    "answer": "Top picks: Brooks Ghost 17, Hoka Clifton 10, ASICS Novablast 5...",
    "organicLinks": [
        { "text": "Shop DICK'S", "domain": "dickssportinggoods.com", "url": "https://www.dickssportinggoods.com/p/..." }
    ],
    "screenshotUrl": "https://api.apify.com/v2/key-value-stores/.../records/ad-...-3-0.png",
    "htmlUrl": "https://api.apify.com/v2/key-value-stores/.../records/ad-...-3-0.html",
    "scrapedAt": "2026-10-02T17:39:41.221Z"
}

turn and message tell you which follow-up triggered the ad. screenshotUrl and htmlUrl are null unless Save ad proof is on.

• adId: the same ad gets the same ID in every chat and run, so count unique ads by adId. openAiAdId, campaignId, and adGroupId are OpenAI's own IDs, filled only when the advertiser's tracking link includes them.
• landingDomain: group by this rather than advertiser. The advertiser name doesn't always match the site (a Beautylish ad can link to goodmolecules.com).
• organicLinks: the links ChatGPT put in the same answer for free, usually the stores it recommends. Compare their domain with landingDomain to see who pays for the spot and who ChatGPT recommends anyway.
• price: brand and category ads (image_card_v2) usually link to a whole store or collection and have no price. When one names a single price in its text, like "now $69.99", that price is used and priceSource is "text". Discounts like "$20 off" and thresholds like "orders over $50" are never taken as the price.
• chatId: ads with the same chatId were shown in one chat. Use it with the CHATS record to get ads per chat.

You can download the dataset as JSON, CSV, Excel, or HTML, or get it through the Apify API.

Summary and chats

Chats without ads aren't in the dataset, so two more records in the Output tab show what happened:

• Summary: for each query, the chats run, adRate (share of chats that showed at least one ad), ads found, and advertisers by ad count. Use adRate to compare queries.
• Chats: one line per chat attempt, with how it ended:

Status	Meaning
done	Ran all its follow-ups.
no-ads	No ad by turn 3, so it was stopped. If its query already showed ads, it runs once more in a new browser, charged once.
blocked	ChatGPT didn't load. Retried, not charged.
failed	Broke mid-chat. Retried, not charged. If it already found ads, they're kept and it isn't retried.
cut	Stopped because the run timeout was near.
skipped	Its query had no ads in 8 chats, so it didn't run. Not charged.

Tips

• Pick topics advertisers pay for. The topic matters more than the wording. Products, travel, software, and services all get ads. ChatGPT shows no ads on health, mental health, or politics, and few on relationships or news.
• Expect empty results. Many chats show no ad, and the same query can return different ads on each run. Run the same queries repeatedly, e.g. on a daily schedule.
• Give it memory. More memory runs more chats at once. With the default 16 GB, a single query at Standard depth (24 chats) finishes in under 10 minutes. Add memory for more queries, but leave at least 1 GB of your account's memory limit free, or the run may find fewer ads.

Limitations

• Most ads are in the US. ChatGPT ads launched in the US. Other countries can be picked with Country, but may return few or no ads.
• Logged-out view. You see the ads ChatGPT shows to logged-out visitors.
• Sporadic ads. ChatGPT doesn't show an ad on every answer, so the scraper has to run many chats to find them. Each chat uses proxies and compute, so each one is charged. We're working on better pricing.
• Residential proxies only. ChatGPT only shows ads to residential IPs, so datacenter proxies don't work.

Pricing

Two things are charged:

Event	Price
Chat	$0.045 for each chat with ChatGPT
Ad	$0.015 for each ad saved to the dataset

Platform usage and residential proxies are included. You don't pay extra for:

• Retries. A chat that gets blocked is run again in a new browser, charged once. So is a chat that shows no ads, once its query has shown ads in another chat.
• Topics with no ads. A query with no ads in 8 chats stops early, and its other chats aren't charged.

Example: one query at Standard depth (24 chats) costs $1.08 for the chats, plus $0.015 per ad. Popular topics often get 2 to 3 ads per chat, so about $1.80 to $2.20 in total. Set Maximum cost per run to cap it. The run stops there and keeps everything found.

Related Actors

• Bing Copilot Ads Scraper: collects the sponsored ads Microsoft Copilot shows in its answers. Run both to compare who advertises on ChatGPT and on Copilot.

• Google AI Overview Ads Scraper: collects the sponsored ads Google shows in its AI Overviews. Run both to compare who advertises on ChatGPT and in Google's AI answers.

Feedback

This scraper is under active development. Found a bug, missing data, or have an idea? Open an issue in the Issues tab.