An AI keyword research tool is only worth trusting if its numbers come from a lookup, not from the model. The setups that hold up pair an AI assistant, which is good at ideas, grouping and intent, with a keyword data source it can call mid-chat, so search volume and difficulty are measured. You can get that from a big suite’s own MCP (Ahrefs lists API and MCP access from its $129/mo Lite plan; Semrush lists MCP on Semrush One Starter and Pro+ and on SEO Classic Pro and Guru) or from a hosted MCP server such as The SEO MCP, free to start with 1,000 credits once you verify your email.
Last updated: October 2026
What changed is where the numbers come from
The label “AI keyword research tool” now covers products that work in completely different ways. The label won’t tell you which one you’re looking at.
What changes everything here isn’t a smarter model. It’s that an assistant can call a data tool in the middle of a conversation, read the result and keep going. Once that works, the model stops being the source of your numbers and becomes the analyst sitting on top of them. That’s a far better job for it.
So the useful question about any AI keyword tool is short: when it shows you a search volume, where did that number come from?
Everything below is a way of answering that.
Three things sold under the same name
A chat assistant on its own. ChatGPT, Claude or Gemini with no data connected. Excellent for brainstorming, clustering and guessing intent. The volumes it adds are plausible-looking estimates, and our post on ChatGPT keyword research with real data explains why a fresh chat can hand you a different figure for the same keyword.
A suite’s keyword database, reached from your assistant. Ahrefs and Semrush both ship their own MCP servers, so on a plan that lists MCP access you can query their data from a chat. Their dashboards are still there for long filtering sessions, and Semrush’s keyword database is one of its real strengths.
A hosted MCP server. One URL you add to your assistant, which gives it a set of SEO data tools with no suite attached. The SEO MCP is one of these: 93 tools across 12 categories, with keyword work spread over the Keywords and Labs categories. Our post on the MCP server for keyword research walks through how those tools chain from seed to SERP.
The second and third can both be a real AI keyword research tool. The first can’t, on its own.
Five checks to run before you trust one
These work on any tool, ours included. Run them in your first session, before a content plan depends on anything it tells you.
1. Can you see which tool produced the number?
Most MCP clients show tool calls inline and let you open the raw input and output. If a volume shows up with no visible call behind it, treat it as a guess. My rule is blunt: a number you can’t trace doesn’t go in a plan. That’s why this check comes first.
2. Ask the same question twice
Open a fresh chat and ask for the volume of the same five keywords. A lookup against the same source should come back the same or very close. A guess often won’t. It takes two minutes, and it’s the quickest way I know to catch a tool that’s improvising.
3. Price a list, not a keyword
Keyword research happens in lists, so the cost that matters is the cost of a list. On The SEO MCP, a Google Ads search volume call is 130 credits for up to 1,000 keywords at Growth Engine rates, per our tools page. Asking about 200 keywords one question at a time means 200 calls. Asking for the whole list means one.
4. Does it cover the whole chain?
Volume alone doesn’t pick a keyword. You also need difficulty, intent and a look at who already ranks, ideally in the same conversation. At Growth Engine rates, The SEO MCP’s catalog lists:
- bulk keyword difficulty, 35 credits per 100 keywords
- search intent classification, 35 credits per 100 keywords
- keyword ideas from a seed, 35 credits per 100 results
- live Google organic results, 8 credits plus 3 per 10 more results
There’s one more measurement worth asking any tool about: how often people ask AI assistants about a keyword. The SEO MCP has an AI search volume tool covering ChatGPT, Gemini and Perplexity, with a 12-month trend, at 29 credits per 100 keywords. Whether it matters depends on your audience, but I’d at least look.
5. How do you pay for it?
Suites charge a monthly plan for the whole suite. Ahrefs Lite is $129/mo with 1,000 credits per user. Semrush’s SEO plan is $139/mo, and Semrush One Starter, which lists MCP, is $199/mo.
The SEO MCP is a monthly plan too. After the free credits, plans start at $99/mo for 50,000 credits, shared across every tool and priced per account with no per-user fees. Full plan details are on the pricing page. Match the plan to how often you actually research keywords, not to the longest feature list.
A first-session test
Here’s how I’d test any AI keyword research tool in one sitting, using checks one and two together.
Start with ideas and let the assistant run free: “Give me 40 keyword ideas around [your topic]. No numbers yet.”
Then measure: “Pull Google search volume for all 40 in one batch, add keyword difficulty and search intent, drop anything under 100 searches a month, and tell me which tool produced each column.”
Now open a second chat and run the measuring prompt again on the same list. Compare the two tables. If the volumes match and every column names a tool, you’ve got a lookup. If they drift and no tool is named, you’ve got a brainstorm with numbers attached.
On The SEO MCP, that measuring step is one volume call, one difficulty block and one intent block, about 200 credits at Growth Engine rates. Rates differ by plan, so once you’re connected, ask the assistant to read the server’s pricing resource for exact costs on yours.
Where The SEO MCP fits, and where it doesn’t
It fits if you already work in Claude, ChatGPT or Cursor and want measured keyword data without opening a dashboard. Setup is the hosted URL plus a browser sign-in with your DFY Hub account; our Claude MCP SEO post covers connecting it to Claude.
It won’t suit everyone. There’s no dashboard, so if you like scrolling a long keyword table with a dozen filters, a suite does that better. Volumes are estimates of demand, not traffic forecasts, and different providers measure differently, so don’t expect them to match an old export. And the assistant 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.
If keyword questions only come up a few times a quarter, you may not need a paid tool at all. Our free keyword research tool guide covers the Google stack. For the wider map, see our guide to AI SEO tools.
Frequently asked questions
What is an AI keyword research tool?
It’s a keyword tool built around an AI assistant that generates, groups and classifies keywords. The ones worth trusting also connect that assistant to a keyword data source, so search volume and difficulty come from a lookup rather than from the model. Without that connection, the volumes an assistant gives you are estimates that can change from one chat to the next.
Can ChatGPT or Claude do keyword research on their own?
They’re good at ideas, clustering and intent. They don’t have direct access to a search volume or difficulty database, so numbers they produce alone are guesses. Connecting an SEO MCP server gives them tools that fetch keyword data when you ask, and most MCP clients show which tool was called, so you can check where each figure came from.
How do I check whether an AI keyword tool uses real data?
Ask which tool produced each number and open the raw result. Then request the same keywords in a fresh chat: a lookup against the same source should come back the same or very close, while a guess often changes. Finally, check that it returns difficulty and intent next to volume, since volume alone won’t tell you whether a keyword is worth targeting.
How much does The SEO MCP cost for keyword research?
You get 1,000 credits once you sign up for a DFY Hub account and verify your email, one time, with no card. Paid plans are monthly: $99 for 50,000 credits, $199 for 200,000 and $399 for 1,000,000. At Growth Engine rates, a search volume call is 130 credits for up to 1,000 keywords. There are no SEO credit top-ups; you upgrade instead.