OpenAI’s public crawler documentation now lists OAI-AdsBot, a bot that will go to pages submitted as ChatGPT adverts to test coverage compliance and assist decide advert relevance.
The entry sits alongside OAI-SearchBot, GPTBot, and ChatGPT-Consumer on OpenAI’s crawler docs web page, bringing the documented bot rely to 4.
OpenAI states that OAI-AdsBot solely visits pages submitted as adverts and that the info it collects isn’t used to coach its generative AI basis fashions.
What The Bot Does
Per OpenAI’s docs, OAI-AdsBot could go to an advert’s touchdown web page after the advert will get submitted. The bot checks whether or not the web page complies with OpenAI’s advert insurance policies. It could additionally use content material from the touchdown web page to assist determine when to point out the advert to ChatGPT customers.
The bot identifies itself with the user-agent string Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko); appropriate; OAI-AdsBot/1.0; +https://openai.com/adsbot.
OAI-SearchBot and GPTBot are each at model 1.3, per OpenAI’s docs. The crawler solely visits pages submitted as advert touchdown pages, not the broader internet.
What The Bot Doesn’t Do
Knowledge collected by OAI-AdsBot isn’t used to coach generative AI basis fashions. That retains OAI-AdsBot out of GPTBot’s territory, which handles coaching knowledge assortment.
It additionally retains OAI-AdsBot separate from OpenAI’s different bots. OAI-SearchBot surfaces content material in ChatGPT search, whereas ChatGPT-Consumer fetches pages throughout user-initiated looking, and OAI-AdsBot is restricted to advert validation.
OAI-SearchBot and GPTBot may be managed independently by means of robots.txt. ChatGPT-Consumer is user-initiated, and the corporate notes that robots.txt guidelines could not apply to it. The OAI-AdsBot entry doesn’t say how the bot treats robots.txt.
No Public IP Checklist But
OpenAI publishes IP vary information for its three earlier bots at openai.com/searchbot.json, openai.com/gptbot.json, and openai.com/chatgpt-user.json. On the time of publication, no equal openai.com/adsbot.json file seems in OpenAI’s docs.
And not using a printed checklist, verifying an actual OAI-AdsBot go to turns into tougher. Consumer-agent strings may be spoofed, and the IP lists offer you a strategy to cross-check for the opposite three OpenAI bots. For OAI-AdsBot, that cross-check isn’t out there.
Why This Issues
OAI-AdsBot has two audiences. Advertisers shopping for placements on ChatGPT want the bot to succeed in their touchdown pages; in any other case, the advert could not validate. Anybody monitoring AI bot exercise in server logs will get a brand new user-agent to observe, one tied to paid stock quite than search or coaching.
Aggressive bot safety by means of Cloudflare, Akamai, or comparable instruments could block OAI-AdsBot earlier than it reaches the web page. That might create validation friction for advertisers who use strict bot-mitigation instruments.
Wanting Forward
ChatGPT’s advert program has moved quick since OpenAI began testing adverts on Feb. 9. As entry opens as much as extra advertisers, OAI-AdsBot site visitors will begin exhibiting up in additional server logs. Look ahead to an eventual IP vary file at openai.com/adsbot.json if OpenAI chooses to publish one. For now, the user-agent string is what it’s a must to work with.
Featured Picture: Blossom Inventory Studio/Shutterstock
