Connect ChatGPT to growResonance
Connect your own ChatGPT to your workspace and let it put your recommendations into practice.
What you need
- Access to the workspace you connect. For the “Read and report back” permission, the Owner or Admin role.
- An active growResonance subscription (a trial is enough).
- A ChatGPT account that allows custom connectors (developer mode must be available).
- In Business and Enterprise workspaces: custom MCP connectors approved by an administrator.
Setup
ChatGPT connects external data sources through the same open standard as Claude: MCP. With the growResonance connection, your own ChatGPT reads the data of your workspace (visibility, prompts, competitors, sources, sentiment) and the recommendations with ready-made artifacts, and keeps working with them right in your chat.
- Turn on developer mode
Open ChatGPT in the browser and go to Settings → Apps → Advanced settings. Turn on developer mode there. In Business and Enterprise workspaces, an administrator has to allow it first (Workspace settings → Permissions and roles → Connected data → Create custom MCP connectors). - Copy the server address
In growResonance, open Settings → Integrations and click “Connect” in the “ChatGPT” row. On the “Connect AI” page, ChatGPT is already selected; click “Connect with ChatGPT”. This copies the server address to your clipboard and opens ChatGPT in a new tab. - Add the connector
In ChatGPT, go to Settings → Connectors and click “Create”. Give it a name, for example growResonance, paste the copied server address and choose OAuth as the authentication. You do not enter a key here: ChatGPT finds the sign-in method by itself. - Confirm access
ChatGPT redirects you to growResonance. Sign in if needed, choose the workspace ChatGPT may see, and confirm. You are then back in ChatGPT and the connector is connected. - Turn it on in the chat and test it
Start a new conversation, click the plus icon, then “More” and “Developer mode”, and turn on the growResonance connector. Then ask for your open recommendations. If a list comes back, the connection works.
What you can do with the connection
With the connection, ChatGPT reads your growResonance analysis. Instead of copying findings out of the dashboard, you ask for them in the chat and have them implemented where your website lives.
- Ask about your own visibility per model, period and location, with the same numbers as in growResonance.
- Compare competitors and cited sources, and see the sentences in which a source appears in the answers.
- Get open recommendations sorted by priority and explained, including why a recommendation comes first.
- Use the ready-made artifact of a recommendation: FAQ JSON-LD, schema markup, meta tags or a draft text, added in the right place.
- Work through the issues from the last Technical Audit, each with affected page, severity and specific fix instructions, filterable by critical, warning and notice.
- Get the full text of a GEO article as Markdown or as finished HTML and move it into your CMS, repository or editor.
- Report back what is done: check off a recommendation or enter the live URL of a published article so the dashboard stays in sync.
What the connection accesses
Access is limited to 17 clearly defined tools. ChatGPT does not get open access to your account but can query and report back exactly the following:
| Tool | Which data | Direction |
|---|---|---|
| Workspace overview | Brand and domain, measured models and measurement frequency, markets, locations, number of competitors, last completed run, active prompts and plan limit, Search Console status. | reads |
| Get visibility | Visibility, share of voice, average position and sentiment of your own brand per model, period and day, with the change from the previous period. | reads |
| List prompts | Your prompts with ID (P-n), status, overall visibility and visibility per model in the last run, trend, mentions and average position. | reads |
| Get a prompt with answers | One prompt with its history and, per model, the answer from the last run: mentioned yes or no, position, brands named, cited sources and answer text. | reads |
| Compare competitors | Ranking of all brands with visibility, share of voice, average position, sentiment, citation rate and visibility per model. | reads |
| List sources | Most cited domains and pages with number of citations, prompts, models, source type and the share of answers that name your brand. | reads |
| Source details | One domain with citation rate, pages, citations per model and day, and the sentence from the answer in which the source appears. | reads |
| Get sentiment | Sentiment toward the brand (0–100), positive, neutral and negative shares, praise and criticism per attribute with verbatim sentences, comparison with competitors. | reads |
| List researched prompts | Prompts from Prompt Research with area, intent, demand, Google and AI search volume, evidence and status. | reads |
| Get Search Console data | Clicks, impressions, CTR and average position from the connected Search Console, plus the strongest queries and pages. Only the numbers, never access to Google. | reads |
| List recommendations | Your open and in-progress tasks from the AEO Strategy: ID (for example A-48), key, type, category, status, priority, title, rationale and the kind of attached artifact. | reads |
| Get a recommendation | A single task with steps, acceptance criteria and affected page, plus, if available, the ready-made building block to use: FAQ JSON-LD, schema markup, meta tags or a draft text. | reads |
| List articles | Your GEO articles: title, slug, status, GEO score, word count, archetype and the live URL, if published. | reads |
| Get article content | The full text of an article as Markdown and as complete HTML, plus the meta description. | reads |
| List audit issues | The findings of the last completed audit run (page speed, AI optimization and SEO), each with severity, affected page URL and fix instructions. | reads |
| Mark a recommendation as done | Sets a task to done, like checking it off in the AEO Strategy. Safe to repeat; dismissed tasks stay dismissed. | writes |
| Report an article as published | Enters the live URL of an article that is already published and sets its status to published. | writes |
What the connection does NOT access
- No credentials: not your password, not the logins of connected accounts (for example Google), not stored API keys.
- No billing data: no invoices, no payment methods, no access to Stripe. From your plan, only the number of allowed prompts.
- No account or team data: no members, no email addresses, no roles, no invitations.
- No data from other workspaces, even if you own several.
- No dismissed recommendations: suggestions you have hidden stay hidden and cannot be brought back through the connection.
- Nothing created, nothing deleted: the connection cannot create or remove prompts, tasks, articles or audit runs.
Example prompts to try
- “How has our visibility developed over the last 30 days?”
- “Show me my open growResonance recommendations.”
- “Summarize the three most important recommendations and explain why they come first.”
- “Get the meta tags from the most important recommendation and write them out for our home page.”
- “Show me the critical issues from my last Technical Audit.”
- “The recommendation is implemented. Mark it as done.”
For developers
- The endpoint speaks Streamable HTTP at https://www.growresonance.ai/api/mcp and responds with JSON.
- Authorization follows OAuth 2.1 with PKCE; the metadata is at /.well-known/oauth-authorization-server and /.well-known/oauth-protected-resource.
- Client registration via Client ID Metadata Document or dynamic registration; public clients without a client secret are supported.
- The server provides domain tools (visibility, sources, recommendations, articles, audit), not the special search and fetch tools that ChatGPT requires for deep research. It is therefore not meant to be used as a research source in Deep Research.
Troubleshooting and FAQ
There is no “Create” option under Connectors.
Then developer mode is not on yet. Go to Settings → Apps → Advanced settings, turn on developer mode and reload the page. In Business and Enterprise workspaces, an administrator also has to allow custom MCP connectors.
After confirming, ChatGPT ends up on the sign-in page again.
This happens if you are not signed in to growResonance in the browser where ChatGPT runs, or if an ad or cookie blocker interrupts the redirect. Sign in to growResonance once in the same browser, open the connector again afterwards and allow redirects while you set it up.
The connector is there, but ChatGPT doesn’t retrieve anything.
The connector has to be turned on in that conversation: plus icon → More → Developer mode → turn on growResonance. If that does not help, name it explicitly in your question, for example “Ask growResonance for my open recommendations”.
How do I revoke access?
Under Settings → Integrations, in the “…” menu of the “ChatGPT” row → “Manage connection”, you find the “Connected apps” list. Clicking “Disconnect” immediately removes ChatGPT’s access to this workspace. You can also delete the connector in ChatGPT itself.
Does this require a specific plan?
The connection is included in all active plans, including the trial. It is read access to data you already see in the dashboard. Anyone with access to the workspace can connect: members with the “Read only” permission, owners and admins also with “Read and report back”. If your subscription expires, access pauses until there is an active plan again; the connection itself stays in place and works again afterwards.
Is there a request limit?
Yes, 60 requests per minute per workspace. Setting up a connection takes two to three requests, each tool call one; even 20 tool calls in a minute use only a third of the limit. The limit protects against endless loops, not against your work. If it is reached, the server responds with “Too many requests”. Waiting one minute is enough.
Does growResonance see what the AI does with the data?
No. The AI runs in your own environment and works there with your permissions. growResonance only logs which tool was queried when and for which workspace, for traceability and abuse protection. Your conversations, your code and your files are not visible to us.