TL;DR
- ChatGPT now suggests Plugins inside the conversation when a Prompt describes a job a Plugin can do. The user doesn't need to search the Plugin directory or know your name.
- Suggestions appear as a short list with a one-click Install. Getting into that list, and near the top of it, decides who gets the customer.
- This is a new organic discovery surface and distribution channel. It can be measured and improved through Agent Tool Optimisation (ATO).
- You can start tracking whether ChatGPT suggests your Plugin, or a competitor's, for the Prompts that matter to you today.
We've gone live with ChatGPT Plugin discovery tracking.
What OpenAI announced about ChatGPT Plugin discovery at DevDay 2026
At DevDay 2026 on September 29, OpenAI shared that 1.2 billion people now use ChatGPT every week, and announced that ChatGPT will now help those users discover Plugins inside the conversation.
Until now, the only way to find and distribute a Plugin was the ChatGPT Plugin directory. Users had to go looking: browse a category, search by name, or already know your product existed. Now ChatGPT surfaces Plugins organically, suggesting them in its reply when a Prompt describes a job a Plugin can do.
For the full set of announcements, read OpenAI's DevDay 2026 recap.
How ChatGPT now suggests Plugins before users go looking
When a Prompt describes a task a Plugin can complete, ChatGPT can now show a “found plugins that could be helpful” Plugin picker directly in its reply, with an Install button beside each suggestion.
This is the first time a Plugin can be discovered organically: by matching what the user is trying to do, not what they typed into a search box.
- The user describes the job. “What booking connectors cover family suites, villas, and resort stays with live availability?” No brand is named.
- ChatGPT works for a few seconds. It searches for travel-capable Plugins and chooses which to suggest. Live availability is only searched later, once a Plugin is installed and connected.
- Three Plugins appear in the reply. Booking.com, Expedia and Tripadvisor, each with a one-line description and a one-click add.
- The answer names them again. The written reply explains what each Plugin covers and links straight through, starting with “Explore Booking.com”.
What happens behind the Plugin picker
The Plugin picker looks like one step, but ChatGPT gets there in three. Finding candidates, choosing which to suggest, and making a Plugin usable are separate jobs, and your Plugin can drop out at each one.
-
1
search_pluginsFinds candidate Plugins
ChatGPT searches its Plugin catalog for Plugins that could handle the request. It searches the catalog, not the public web, and it returns a bounded set of candidates, not every Plugin that exists.
See what this step does
- Matches concise provider, product and capability terms, such as “hotel booking” or a Brand name, and can match several terms at once.
- Results are explicitly not exhaustive: a Plugin that could do the job may still not come back as a candidate.
- What it can match is what your Plugin says about itself: its name, description and capabilities.
-
2
suggest_pluginsChooses what goes in the Plugin picker
From those candidates, ChatGPT selects which Plugins to put in the install suggestion the user sees: the “found plugins that could be helpful” Plugin picker.
See what this step does
- Picks specific Plugin IDs from the candidate set for the suggestion.
- The shortlist depends on the Prompt and on the user, so it is not guaranteed to be the same for everyone, or every time.
- Being a candidate is not enough: this is where you win or lose the slot.
-
3
Install and connectMakes the Plugin usable
The user installs one suggested Plugin, several, or none. Installing and connecting their account with your service are separate steps, and only after both can your Plugin do real work, such as searching live availability.
See what this step does
- Installation adds the Plugin to the user's ChatGPT.
- Connection authorises the user's account with the external app, where your service requires it.
- Only then does ChatGPT call your Plugin's tools to complete the task.
Why the recommendation is the new search result
Search gave users ten blue links. The Plugin directory gave them a category page. ChatGPT now goes straight from the Prompt to the Plugins that can complete the task. Your Plugin doesn't need to rank on a page anymore. It needs to win the recommendation.
The user goes looking.
- User must already know a Plugin exists
- Search by name or browse a category
- Discovery happens outside the task
- One ranking per category, the same for everyone
ChatGPT brings the Plugin to them.
- User describes the outcome they want
- ChatGPT matches intent to Plugin capabilities
- Discovery happens inside the task
- A shortlist that depends on the Prompt and the user
-
01 · Your user asks
Plan our launch so owners update their own workstreams and risks get flagged.
The user describes the outcome. They don't need to know which Plugin can deliver it.
-
02 · ChatGPT shortlists
ChatGPT searches a bounded set of candidate Plugins and selects a shortlist to show. ATO improves this decision.
-
03 · The user installs
The user installs one, several or none, then connects it to complete the task. Plugins that aren't chosen lose that moment.
Anatomy of a ChatGPT Plugin suggestion: five parts that decide installs
Being shown is step one. The Plugin picker gives the user a few seconds and a handful of words to choose. Here's what's doing the work.
- The trigger. The Plugin picker only appears when ChatGPT judges that a Plugin would complete the task better than a text answer. No trigger, no picker. Your Prompts need to describe work, not questions.
- Slot order. The first slot is read first. In our runs, position is an observed prominence signal that works like a rank, and it can change with the wording of the Prompt.
- Name and icon. In our captures, recognition seems to break ties. When capabilities look similar, a known brand is the easy click, so smaller Plugins have to win on specificity.
- The one-line description. The only copy the user reads before deciding. Compare “Create, search, update docs” (features) with “Turn chats into actions” (an outcome). Neither mentions launch plans or risk flags.
- Install / Not now. The conversion event. Then look at the reply: ChatGPT names the Plugins again in its answer and explains the setup. Getting mentioned there reinforces the pick.
An annotated ChatGPT Plugin picker suggests Coda (“Create, search, update docs”), Asana (“Turn chats into actions”) and ClickUp (“Run your projects from ChatGPT”), with numbered markers matching the five parts below.
AEO vs ATO: how Plugins reach the ChatGPT Plugin picker
The Plugin picker creates a new discovery layer: a high-intent surface where relevant companies are found, compared and chosen at the moment a user is ready to act.
For each Prompt, ChatGPT searches a bounded set of candidate Plugins, selects which make the Plugin picker, and shows them in an order. That search runs over ChatGPT's Plugin catalog, not the public web, so AEO is not the direct mechanism for this surface. It is a different contest from being cited in an answer, and it needs a different discipline.
Answer Engine Optimisation (AEO) improves how accurately a company is represented, cited and recommended when an AI system answers a question. It works on the content and sources answer engines retrieve and trust. It doesn't change whether an agent finds, ranks or picks a callable tool.
Agent Tool Optimisation (ATO) improves whether a product's callable tools are found, picked and positioned when an AI agent decides how to complete a user's request. It works on the names, descriptions, schemas, metadata and documentation agents use to choose between tools. The term was coined by AgentDiscoverability.com.
AEO wins the citation.
Helps information get found, trusted and cited in an answer.
ATO wins the invocation.
Helps a callable Plugin get discovered, selected, installed and used.
- 1UnderstoodChatGPT can tell what your Plugin does and which tasks it completes.
- 2MatchedYour capabilities fit the Prompt better than nearby alternatives.
- 3SurfacedYou earn a slot in the Plugin picker shown inside the conversation.
- 4InstalledNew since launch: the user picks you over the other slots.
AEO vs ATO at a glance
Both matter. They win different moments.
| AEO | ATO | |
|---|---|---|
| The user's moment | Asking a question | Trying to get a job done |
| What you win | A citation or mention in the answer | A slot in the Plugin picker, then the install and the call |
| What you optimise | Web content, sources, reviews, entity data | Plugin name and description, tool names, tool descriptions, schemas, metadata |
| Who decides | The answer engine, choosing sources to quote | The agent, choosing tools to use |
| How it's measured | Mentions, citations, share of voice | Share of suggestion and slot position, measured through repeated runs (ChatGPT does not report them) |
How to get your ChatGPT Plugin suggested: measure, diagnose, improve, verify
Optimising for ChatGPT discovery is a measured cycle, not a one-off rewrite. Run the Prompts your Plugin should win, find where visibility is lost, make focused improvements, then measure the same Prompts again.
Step 1 · Measure Run your priority Prompts repeatedly and measure share of suggestion, slot position, and which competitors appear. ChatGPT does not report these; repeated runs are how you see them.Fix two Prompt sets and keep them unchanged: the high-value intents you most want to win, and a wider category set to benchmark against competitors. Run each Prompt many times, because one run is a sample, not a result.
Step 2 · Diagnose Find where visibility is lost: no Plugin picker at all, not in the picker, or in the picker but outranked by another Plugin.Each loss points somewhere different. Not a candidate means ChatGPT's search isn't matching your Plugin to the task. A candidate that is never suggested means your fit isn't clear enough. Suggested but outranked means a competitor explains the same job better.
Step 3 · Improve Make focused changes to the signals ChatGPT reads: your Plugin name and description, tool names, tool descriptions, schemas and metadata.Start with the highest-impact gap rather than rewriting everything, and record exactly what changed and when it went live, so its effect can be measured cleanly.
Step 4 · Verify Run the same Prompts again and confirm the outcome improved. Keep what worked, then repeat.Compare against your baseline across repeated runs. One good result is not proof: look for a consistent lift, without losing Prompts you were already winning.
Continuous, not one-off. Suggestions shift as ChatGPT, your competitors and your own Plugin change. Each new measurement shows whether your Plugin is getting easier to find and choose.
Case study: organic discovery is already changing how brands get found
See how some brands are already getting their Plugins and Connectors organically discovered.
ChatGPT Plugin discovery FAQ
Answers for product, growth and engineering teams whose Plugin now competes for ChatGPT's suggestions.
What is ChatGPT Plugin organic discovery?
It is ChatGPT suggesting Plugins inside the conversation. When a Prompt describes a job a Plugin can do, ChatGPT shows a short “found plugins that could be helpful” list in its reply, each with a one-click Install. The user does not need to search the Plugin directory or know your Brand name.
What did OpenAI announce about Plugins at DevDay 2026?
At DevDay 2026, OpenAI shared that 1.2 billion people use ChatGPT every week and announced that ChatGPT will recommend relevant Plugins during conversations, alongside the Plugin directory. See OpenAI's DevDay 2026 recap for the full list of announcements.
How does ChatGPT decide which Plugins to suggest?
In three steps. A search tool, search_plugins, finds a bounded, non-exhaustive set of candidates from ChatGPT's Plugin catalog by matching provider, product and capability terms. A second step, suggest_plugins, picks which of those go in the Plugin picker. The user then installs and connects. Your Plugin's name, description and capabilities are what the search can match; signals such as latency, errors or reviews are not documented, so treat them as hypotheses.
How do I get my ChatGPT Plugin suggested in conversations?
Treat it as a measured cycle. Measure how often ChatGPT suggests your Plugin for your priority Prompts, diagnose where you lose (no Plugin picker, not in the picker, or outranked), improve the signals ChatGPT reads, then verify with the same Prompts. This practice is Agent Tool Optimisation (ATO).
Is Plugin discovery the same as SEO or AEO?
No. SEO and Answer Engine Optimisation (AEO) help content get found, cited or mentioned in search results and AI answers. Plugin suggestions come from a search over ChatGPT's Plugin catalog, not the public web, so AEO is not the direct mechanism here. Agent Tool Optimisation (ATO) is: it works on your Plugin's names, descriptions, schemas and metadata.
Do I still need a listing in the ChatGPT Plugin directory?
Yes. The ChatGPT Plugin directory is still where users browse and install Plugins. Organic suggestions add a second way to be found: inside the conversation, at the moment of intent. New to Plugins? Start with what a ChatGPT Plugin is and how to build one.
How many Plugins does ChatGPT suggest at once?
A small shortlist. Every example we have captured so far showed three Plugins, and there is no second page. Position in that list matters because the first slot is read first, and “Not now” dismisses the whole Plugin picker.
Does Claude suggest Connectors in the same way?
Yes. Claude can also surface relevant Connectors inside a conversation for action Prompts. See how to get your Claude Connector organically discovered, including how DirectBooker moved from a 31.5% to a 100% pick rate.
How can I track whether ChatGPT suggests my Plugin?
AgentDiscoverability.com runs your priority Prompts in ChatGPT repeatedly and tracks whether your Plugin is suggested, its slot position, which competitors appear and which MCP changes would improve discovery. Sign up to start tracking.
Start tracking your ChatGPT Plugin's organic discovery today.
- Your Plugin tracked in ChatGPT
- Competitors in your slot
- Share of suggestion for your Prompts
- Recommended MCP changes to improve discovery
Related reading
- How to get your Claude Connector discovered organically. A four-stage process to measure, diagnose, improve and verify Claude discovery.
- What is a ChatGPT Plugin?. The Plugin Directory explained, and what a Plugin contains.
- ChatGPT Plugin discoverability: the complete guide. Metadata, prompt testing and optimization for ChatGPT Plugins.
- MCP Directory: every ChatGPT Plugin and Claude Connector. Search and compare listings, ranked and updated daily.
- Check your Discoverability Score. See how well agents find and pick your product.