What is the Keyword Research API?
The Keyword Research API takes one seed phrase, such as “standing desk”, and returns the longer searches people type around it: “electric standing desk”, “desktop converter standing desk”, “sit and standing desk”, 17,795 ideas in all in our test. Every idea comes with the numbers you need to decide whether it deserves a page or an ad group: monthly searches, a 12-month series, keyword difficulty, cost per click and the intent behind the search.
It covers the discovery step of keyword research. You start from a topic, get the long tail, and shortlist the queries worth a closer look before anyone writes a brief.
You ask
Find 20 keyword ideas for “standing desk” in the US, sorted by search volume, with difficulty, CPC and intent.
It runs
From your AI assistant over MCP or from your code over REST
You get
- Keyword
- Demand
- Advertising
- Ranking
- Results page
- Plan contentTurn a product category into article and landing-page topics, grouped by what the searcher wants to do.
- Find easier targetsSort by keyword difficulty and spot long-tail queries with real demand and fewer strong pages competing.
- Size a nicheAdd up demand across distinct ideas to see whether a topic can carry a whole content cluster.
- Price paid searchCompare CPC and top-of-page bids across themes before you build campaigns around them.
- Hand it to an agentLet Claude or ChatGPT pull the ideas, cluster them and draft briefs in the same conversation.
What data does the keyword research API return?
Each idea comes back as one record. These are the fields our test run returned, described in plain words.
Keyword
- Keyword text
- Word count
- Synonym group it belongs to
- Detected language
Demand
- Average monthly searches
- Searches for each of the last 12 months
- Change over a month, a quarter and a year
Advertising
- Cost per click
- Low and high top-of-page bid
- Advertiser competition, from 0 to 1 and as low, medium or high
- Product and service category codes
Ranking
- Keyword difficulty, from 0 to 100
- Main search intent and secondary intents
- Average backlinks and referring domains of the pages that rank
- Date the metrics were last updated
Results page
- Result types on the page, such as AI Overview, popular products and ads
- Number of results
- Link to the live results page
Results-page details are optional and arrive when you switch them on. Already have a list of terms and only need numbers? Keyword Metrics scores a known list, and Keyword Search Volume API for Google Ads returns Google Ads volume and CPC for up to 1,000 terms.
Sample output from a real keyword run
The table at the top of this page is a real run from 16 November 2026: seed “standing desk”, United States, English, 20 ideas sorted by search volume. The run reported 17,795 ideas in total for this seed, so the 20 you see are the head of a much longer list. Below the chart is one idea with every field filled in.
One result in full
One idea in full: electric standing desk
- Average monthly searches
- 22,200
- Highest month
- 49,500 (March 2026)
- Lowest months
- 9,900 (September and October 2026)
- Change
- 0% month, -33% quarter, -55% year
- Cost per click
- $2.92
- Top-of-page bid
- $0.46 to $5.00
- Advertiser competition
- High (1.0)
- Keyword difficulty
- 51 of 100
- Search intent
- Transactional
- Avg. referring domains of ranking pages
- 387
- Metrics updated
- 12 Nov 2026
What this sample tells you
10of 20
10 of 20 ideas report 135,000 monthly searches, the same figure as “standing desk” itself. They are rewordings such as “desk for standing” and “standing at a desk”, so count that group once instead of adding the numbers up.
10of 20
Only 10 of 20 contain the exact phrase “standing desk”. The others use both words in a different order, like “desk standing converter”. Turn on exact match when word order matters to you.
Difficulty runs from 5 to 51. Four ideas score 11 or lower with 18,100 to 22,200 searches a month: two brand-name queries and two converter phrasings.
17of 20
17 of 20 ideas are transactional and 3 are informational. All 20 carry high advertiser competition, with CPC from $2.24 to $5.78.
How to research keywords with your AI assistant
Connect AI/ML API
Click Try it now, pick your client and add the MCP server. You sign in with your AI/ML API account once.
Give a seed and a market
Name the seed phrase, the country and the language, and say how many ideas you want back.
Read the shortlist
Your assistant returns each idea with volume, difficulty, CPC and intent, sorted the way you asked.
Turn ideas into briefs
Ask it to merge rewordings, drop off-topic ideas, group the rest by intent and outline a page for the strongest cluster.
Prompts to start with
Use the AI/ML API MCP Keyword Research API for Seed Ideas tool. Find keyword ideas that contain “standing desk” in the US and show search volume, keyword difficulty, CPC and intent in a table, sorted by volume.
Use the AI/ML API MCP Keyword Research API for Seed Ideas tool. Give me 50 keyword ideas for “trail running shoes” in the United Kingdom with keyword difficulty under 30, sorted by search volume.
Use the AI/ML API MCP Keyword Research API for Seed Ideas tool. Find informational keyword ideas for “espresso machine” in Germany, in German, and group them into article topics.
Use the AI/ML API MCP Keyword Research API for Seed Ideas tool. Pull 100 ideas for “standing desk”, merge rewordings that share one volume figure, and list the 10 groups with the most demand.
Settings that shape your keyword list
In chat you describe what you want and your assistant fills these in. In code you pass them yourself.
| Setting | What it controls | Tip |
|---|---|---|
| Seed keyword | The phrase every idea is built around | Use the words buyers type: “standing desk”, not “ergonomic office furniture” |
| Location | Country or region for the metrics, by name or code | Run each market separately, since volumes and CPC differ by country |
| Language | Language of the searches, such as English or German | Match the language people actually search in for that country |
| Number of ideas | 1 to 1,000 rows per run, 100 by default, with an offset for the next page | Start with 20 to 50 to check relevance, then page through |
| Exact match | Keep only ideas that contain the seed exactly as typed | Turn it on when word order changes the meaning |
| Filters | Up to 8 rules on any metric, such as volume above 1,000 or difficulty below 30 | Filter in the request instead of downloading thousands of rows |
| Sort order | Up to 3 sort rules, for example search volume, highest first | Sort by difficulty for quick wins and by volume for big themes |
| Ignore synonyms | Drop near-identical rewordings from the list | Useful when one wording floods the list with variants |
| Seed row and results-page data | Add a row for the seed, and the result types Google shows for each idea | Switch on when you want to know what a page competes with before writing |
Monthly searches are 12-month averages for the market you choose. A seasonal product can look bigger or smaller than it is this month, so check the monthly series before you plan around one number.
Ways teams use keyword research data
Content calendars
Expand each product category into ideas, keep the informational ones and schedule a guide for every cluster.
KeywordIntentSearch volume
Low-difficulty targets
Filter for difficulty under 20 and volume over 1,000 to shortlist topics worth testing on a newer site.
DifficultySearch volume
Paid search planning
Compare CPC and top-of-page bids across themes before you set budgets or split ad groups.
CPCBidsCompetition
Seasonal timing
Use the 12 monthly figures to publish or raise bids ahead of the peak, not after it.
Monthly searchesChange
Product and naming research
See which brands, models and features people pair with a category, such as converters or electric frames.
KeywordSearch volume
Briefs with an agent
Have your assistant pull ideas, merge rewordings, check the results page and draft the brief in one chat.
Any field
Keyword data API for your code
In your code
REST · your AI/ML API key, JSON records back
The same tool runs over REST with your AI/ML API key. Send a seed, a location and a language and you get JSON with one record per idea: the keyword, monthly searches, the 12-month series, CPC, bids, competition, difficulty and intent. Put filters and sort rules in the request so you only download the rows you will use.
Use MCP when a person researches in a chat and REST when a script refreshes keyword lists for a dashboard or a content plan. Both use one AI/ML API account and one balance. Start with the Toolbox REST quickstart.
Other ways to get keyword ideas
You can build the same list several ways. The right one depends on how many seeds you research and where the rest of the work happens.
| Google Keyword Planner by hand | Paid SEO suite in the browser | Build on an official ads API | ||
|---|---|---|---|---|
| Setup | Connect once with your AI/ML API account | Google Ads account, then copy or export | Monthly subscription and seats | Developer token approval, OAuth and your own code |
| Inside an AI chat | Yes. Over MCP | No. No | Only through exports | Only if you build it |
| Best for | Repeat research from chat or code on one balance | Occasional checks of ad demand and bids | Teams that work in one dashboard all day | Products that need first-party ad data |
Pricing
Pay as you goBilled per run from your AI/ML API balance
- No subscription
- No seat fee
- One balance for every tool
Each run is billed from your AI/ML API balance, with no subscription or seat fee. The price grows with the number of ideas you ask for, so a 20-idea check costs less than a 1,000-idea export. Our 20-idea test run cost under two cents. Check the current price in your dashboard before large runs.
FAQ
How accurate is the search volume?
Monthly searches are estimates for the market you pick, rounded into bands such as 18,100, 22,200 and 135,000. Use them to compare ideas against each other, not as exact counts of searches.
Why do several ideas show exactly the same volume?
Close rewordings of one search are reported with one shared figure. In our test, 10 of 20 ideas showed 135,000, the same as the seed. Treat them as one topic and count the number once.
Is keyword difficulty the same as advertiser competition?
No. Keyword difficulty estimates how hard it is to rank in organic results, on a 0 to 100 scale. Advertiser competition describes how crowded the Google Ads auction is for the term. In our run every idea had high competition while difficulty ranged from 5 to 51.
How is this different from Related Keywords?
This tool returns ideas built on the words of your seed. Related Keywords finds adjacent searches that may not contain the seed at all, which helps when your seed is too narrow.
How is this different from Google Keyword Planner?
Keyword Planner lives inside a Google Ads account and is built for advertisers. The Keyword Research API also returns organic metrics such as keyword difficulty and intent, runs on your AI/ML API account and works inside AI assistants over MCP.
How many keyword ideas can I get from one seed?
Up to 1,000 per run, and you can page through the rest with an offset. Our “standing desk” seed had 17,795 ideas in the United States, so filter by volume or difficulty before you page through everything.
Can I use it from Claude, ChatGPT or Cursor?
Yes. Add the AI/ML API MCP server once and ask in plain English. Hermes, OpenClaw and other MCP clients work the same way.
How fresh is the data?
Each idea carries the date its metrics were last updated. In our 16 November run those dates ran from 17 September to 13 November 2026, and the monthly series ended with October 2026.
Can I export the ideas to CSV or Google Sheets?
Ask your assistant to return the table as CSV, or write the JSON from the REST response to a file or a sheet in your code.
Research your first seed
Connect once, then ask for keyword ideas in plain English.