ChatGPT keyword research with real data needs a data source, because ChatGPT on its own has no search volume or difficulty database. Ask it for volumes and you get estimates that look exact, and with web search on, maybe a figure lifted from someone else’s blog post. There are two ways to get real numbers into the chat. You can paste or upload an export from a keyword tool, or connect an SEO MCP server so ChatGPT calls live keyword tools itself. The SEO MCP connects through Developer mode in ChatGPT on the web, with one hosted URL and a browser sign-in.
Last updated: October 2026
The difference hardly anyone checks
Most people know ChatGPT can brainstorm keywords. Fewer notice that the numbers next to those keywords aren’t measurements.
A language model writes the most plausible answer. “Around 2,400 searches a month” is a plausible-looking answer, so that’s what you get, often with a difficulty score attached that came from nowhere. Ask the same question in a fresh chat and the figures can change. Nothing about the keyword changed. The guess did. That’s why a list built in one chat can disagree with a list built in the next.
That gap matters more than it seems. A content plan built on guessed volumes ranks topics in the wrong order, and you don’t find out until months of writing go to terms nobody searches. My rule is blunt: never put a volume ChatGPT produced on its own into a content plan. Use it for ideas, grouping and intent, and get the numbers from somewhere that measures them.
Two ways to get real numbers into ChatGPT
1. Paste or upload an export. Run your keywords through a keyword tool, export the results, and give the file to ChatGPT to sort, cluster and summarize. Pasting works on any ChatGPT plan, including free, and it’s how most people start. The downside is that ChatGPT only sees the slice you exported. When it needs one more metric or a follow-up keyword, you’re back in the other tool, exporting again.
2. Connect an MCP server. With an SEO MCP server connected, ChatGPT can request the data itself mid-chat. You describe what you want, it calls the tools, reads the results and answers. If it needs more, it makes another call. Our ChatGPT SEO MCP post covers what else that connection can do, and Does ChatGPT Mention My Brand? covers checking what ChatGPT says about you.
If you only research keywords a couple of times a year, the first route is fine. If keyword questions come up every week, the second saves the round trips. (For free tools that make the first route cheap, see our post on the free keyword research tool options.)
Connecting The SEO MCP to ChatGPT
These steps match our install guide for ChatGPT:
- In ChatGPT on the web, open Settings → Security and login and turn on Developer mode.
- Go to Plugins, select the + button, and create an app for a remote MCP server.
- Enter the server URL:
https://mcp.yourdfyhub.com/seo
OAuth sign-in then opens in your browser, and you sign in with your DFY Hub account. There’s no API key to paste and nothing to install.
The limitation, up front: Developer mode is on Plus, Pro, Business, Enterprise and Education plans, in the web app. On Business and Enterprise workspaces an admin may need to allow it first. If you’re on free ChatGPT, the export route is the one open to you.
What “real data” covers once you’re connected
These are the keyword tools ChatGPT can call, with costs from our tools page at Growth Engine rates:
| What you ask for | Cost on Growth Engine |
|---|---|
| Google Ads search volume | 130 credits per call, up to 1,000 keywords |
| Bing search volume | 130 credits per call |
| Google Trends data | 16 credits |
| Keyword overview with metrics | 18 credits |
| Keyword ideas from a seed | 35 credits per 100 results |
| Bulk keyword difficulty | 35 credits per 100 keywords |
| Search intent classification | 35 credits per 100 keywords |
| AI search volume (ChatGPT, Gemini, Perplexity) | 29 credits per 100 keywords |
Local Dominator uses fewer credits per call and Full Stack uses more. Once connected, you can ask ChatGPT to read the server’s pricing resource for exact costs on your plan. To confirm each number came from one of these calls, see the checks in our ai keyword research tool post.
Batching is where the savings are. One search volume call covers up to 1,000 keywords, so check a whole list at once instead of one keyword per question. Our MCP server for keyword research post goes deeper on chaining these tools from seed to SERP.
A prompt that swaps guesses for measurements
Here’s the workflow I’d use, and it keeps ChatGPT doing what it’s good at.
Start with ideas. Ask for 40 keyword ideas around your topic and let it brainstorm freely. No numbers yet.
Then measure:
“Take those 40 keywords. Pull Google search volume for all of them in one batch, add keyword difficulty, and classify search intent. Drop anything under 100 searches a month and sort the rest by volume.”
That’s one volume call, one difficulty block and one intent block: about 200 credits at Growth Engine rates. Then ask it to group the survivors into topics and draft a brief for the top one. The grouping and the brief are ChatGPT’s strengths. The numbers under them are now measured.
One habit worth keeping: ask ChatGPT which tool produced each number. Tool names don’t always show up unless you ask, and asking is a quick way to confirm a figure came from a call and not from the model.
What real data still won’t fix
Google Ads search volume is an estimate of demand, not a traffic forecast. It doesn’t tell you how many of those searches click through, so use it to rank keywords against each other.
Providers also measure differently. A volume from one source won’t exactly match another, so don’t judge a tool by whether it agrees with an old export.
And ChatGPT still does the filtering and summarizing, which means it can get that step wrong. When a number decides what you build, open the raw result before you commit.
For the wider view of SEO work inside assistants, see our guide to AI SEO tools.
What it costs
You get 1,000 free credits when you sign up and verify your email, with no credit card. Paid plans are $99/month for 50,000 credits (Local Dominator), $199/month for 200,000 (Growth Engine) and $399/month for 1,000,000 (Full Stack). Every plan includes all 93 tools. Credits reset monthly and don’t roll over, and there are no SEO credit top-ups: if you run out, upgrading adds the new plan’s credits right away. Details are on the pricing page.
Frequently asked questions
Can ChatGPT do keyword research with real search volume?
Not on its own. ChatGPT doesn’t have direct access to a search volume or difficulty database, so figures it produces by itself are estimates. You can paste in an export from a keyword tool, or connect an SEO MCP server such as The SEO MCP so ChatGPT calls live keyword tools during the chat and answers with measured numbers.
Why does ChatGPT give different search volumes each time I ask?
Because without a data source, it’s generating a plausible number, not looking one up. A fresh chat can produce a different guess for the same keyword. Connecting a keyword data source fixes this, since the figure then comes from a tool call. Asking ChatGPT which tool produced a number is a quick way to check.
How do I connect The SEO MCP to ChatGPT?
In ChatGPT on the web, open Settings, then Security and login, and turn on Developer mode. Go to Plugins, select the plus button, create an app for a remote MCP server, and enter the server URL. Sign-in opens in your browser. Developer mode is on Plus, Pro, Business, Enterprise and Education plans.
Does this work on the free ChatGPT plan?
Not the MCP connection. Developer mode, which you need to add a custom MCP server, is on ChatGPT’s Plus, Pro, Business, Enterprise and Education plans. On free ChatGPT you can still get real data in by exporting keyword data from another tool and pasting it into the chat for ChatGPT to sort, cluster and analyze.