OpenAI lets ChatGPT users @ custom GPTs directly into chats

Jumping between different custom GPTs used to feel like walking in and out of isolated chat rooms. If you wanted to plan a trip, generate code, and make a poster, you had to start three separate conversations and repeat your context every single time.

OpenAI is finally fixing that friction. ChatGPT paid subscribers can now bring any custom GPT directly into an ongoing conversation just by typing "@" and picking it from a list.

Tagging in specialists on the fly

The biggest upgrade here is context sharing. The chosen GPT immediately understands the full history of your current chat, meaning you don't have to re-explain yourself.

You can basically tag different bots in and out as your needs evolve. For example, you can brainstorm with standard ChatGPT, ping @AllTrails to find a hiking route, and then call in @Canva to draft a flyer without leaving the thread.

Building these apps still requires zero coding, ranging from simple prompt wrappers to integrations like Khan Academy's code tutor. OpenAI's eventual plan is to let creators monetize their custom bots, but they have a much bigger hurdle to clear first: actual usage.

Low traffic and moderation headaches

The push to make GPTs more visible comes at a critical time. Despite the recent launch of the GPT Store, users simply aren't showing up.

According to data from analytics firm Similarweb, custom GPTs account for a mere 2.7% of ChatGPT's worldwide web traffic. Even worse, that traffic has been slipping month over month since November.

On top of that, OpenAI is wrestling with a chaotic store catalog. During its debut week, the store was flooded with romantically and sexually suggestive bots, directly violating company policies.

Developers also rolled out unauthorized political campaigning tools, including a bot impersonating U.S. presidential candidate Dean Phillips. While OpenAI relies on a mix of human and automated reviews to ban rogue apps, keeping the catalog clean will only get harder as volume grows.

Why it matters

For regular ChatGPT Plus users, this is the first time custom GPTs actually fit into a normal daily workflow. The standalone store model felt clunky, but summoning specialized bots inside a single thread makes multi-step research and content creation worth testing right now.

For the broader AI ecosystem, OpenAI is borrowing the best playbook from Slack and Discord. Moving away from a static storefront toward conversational mentions is a smart move that competitors will likely scramble to copy.

However, we think the fundamental problem isn't just discoverability. A 2.7% traffic share indicates that most custom GPTs offer very little moat or genuine utility. If OpenAI cannot clean up low-quality spam and incentivize truly powerful agentic workflows, lowering the interaction barrier might just expose users to underwhelming tools even faster.