How to Use AI for Keyword Research (Without Trusting Made-Up Numbers)

The short answer: use AI for keyword research as an idea generator, not as a data source. Ask a chatbot such as ChatGPT or Claude for the questions beginners type about your topic, then check every phrase with free Google tools (autocomplete, Keyword Planner, Trends and Search Console) before you write a word. A chatbot’s search numbers are guesses; Google’s tools show what people really search.

You type a question into ChatGPT. “Give me 20 keywords for my new blog.” In ten seconds, you get a neat list, with search numbers next to every line.

It feels like magic. It also feels like you just skipped the hardest part of blogging.

Here is the catch. Many of those numbers are made up. The tool is not lying on purpose. It just does not have the data, so it guesses, and it guesses in a very confident voice. A new blogger who trusts those guesses can spend weeks writing about words nobody searches for.

Good news: AI is still really useful for keyword research. You just need to give it the right job. In this guide I will show you what AI can do well, what it cannot do, and a simple routine you can follow this week, even if you have never heard the word “keyword” before today.

What Is a Keyword, in Plain Words?

A keyword is the words a person types into Google. “How to start a blog.” “Cheap standing desk.” “Why is my sourdough flat.” That is it.

Keyword research means finding out which of those phrases people really type, and picking the ones your new site has a fair chance to show up for. If you skip this step, you are guessing which articles to write. If you do it, you are writing things people already look for.

If you are still choosing what your blog is about, start with how to choose a profitable blog niche first. Keyword research works much better once you know your topic.

Can AI Do Keyword Research on Its Own?

No. AI is good at ideas and grouping, but it cannot see Google’s search data, so it cannot tell you how many people search a phrase or how hard it is to rank.

Think of an AI chatbot like a smart friend who read a huge amount of the internet. That friend is great at ideas. That friend is not a spreadsheet of Google data.

Where AI helps

• Brainstorming. It can turn one topic into dozens of related questions, in seconds.
• Grouping. You can paste a messy list and ask it to sort the phrases into themes. Each theme can become one article.
• Spotting intent. It can tell you if a phrase sounds like someone wants to learn, compare, or buy.
• Wording. It can show you the different ways a beginner might ask the same question.

Where AI fails

• Search volume. A chatbot has no live access to Google’s numbers. SpyFu, an SEO tool company, tested this. ChatGPT guessed a phrase got 5,000 to 10,000 searches a month, while real data showed roughly 700 to 800. That is a gap of about ten times.
• Difficulty. It cannot tell you how hard it is to rank, because it cannot see who ranks today.
• Fresh facts. It may describe tools or trends that changed since it was trained.
• Made-up keywords. Sometimes it invents phrases that look real but nobody types.

Search Engine Land reached the same view in its own ChatGPT keyword guide. It calls the chatbot a brainstorming helper that should be paired with real SEO tools for checking. I agree with that. Use AI to get ideas, and use real data to decide.

Which Free Tools Show Real Search Numbers?

Three free Google tools cover the basics: Keyword Planner for search ranges, Google Trends for direction, and Search Console for the phrases that already bring people to your site.

A note on links: I’m not an affiliate for any of the brands below, so nothing on this page pays me either way.

You do not need to pay for anything to start. These three tools are free to use, and they show you what people really search.

• Google Keyword Planner. This one comes from Google itself. Google says you need to finish your Google Ads account setup, which includes billing details, before you can use the basic features. You do not have to run ads. Rules change, so read Google’s own page for the current steps. Without active ad spend, the numbers often show as wide ranges, not exact counts. That is fine for a beginner.
• Google Trends. It shows whether interest in a phrase is rising, falling, or flat. It does not give search counts. It does let you compare two phrases side by side.
• Google Search Console. Once your site has a few weeks of traffic, this shows the exact phrases that already bring people to you. For a new blogger, that is real gold, and it costs nothing.

Paid tools such as Semrush and Ahrefs also exist. They have free tiers or trials, but their limits and prices change often, so check their own pages before you count on them. You can do a lot without them.

How Do I Do Keyword Research With AI? A 6-Step Routine

Here is the routine I would follow if I were starting from zero. It uses AI for what it is good at and real data for the rest.

Step 1: Write down your topic in one sentence

Be specific. “Budget travel” is too wide. “Weekend trips under $300 for people with a day job” is something you can work with.

Step 2: Ask AI for questions, not numbers

Beginners type “give me keywords with search numbers.” Skip that. Ask for the questions a real person would type. Try this:

I run a beginner blog about [your topic]. List 30 questions a complete beginner might type into Google before they know the right words. Do not give me search volumes. Group the questions by what the person wants to do: learn, compare, fix a problem, or buy.

The line “do not give me search volumes” matters. It stops the tool from handing you pretend numbers.

Step 3: Ask for the “long” versions

Short phrases are crowded. Longer ones are often quieter. “Keyword research” is a hard fight. “How to do keyword research for a new blog with no budget” is a smaller room with fewer people in it. Ask AI to stretch each of your strongest ideas into three longer versions. If you want more prompts you can reuse, grab the ones in 25 AI prompts for starting an online business.

Step 4: Check the list in Google itself

Type each phrase into Google and look at three things:

• Autocomplete. The words Google adds while you type are things real people searched.
• “People also ask.” This box gives you more real questions to write about.
• The results page. Who ranks? If page one is full of huge, famous sites, a brand-new blog will struggle. If you see small blogs and forum threads, there may be room for you.

Step 5: Add real numbers

Put your 10 to 15 strongest phrases into Keyword Planner and look at the ranges. Then check the shape in Google Trends. You are not hunting for the biggest number. You are looking for phrases with some steady interest and results that do not look impossible.

This is also the step where the AI guess gets fixed. If the chatbot sounded sure about a phrase and Keyword Planner shows almost nothing, trust Keyword Planner.

Step 6: Pick one phrase per article

Choose one main phrase for each post. Put it in the title, the first paragraph, and one or two headings. Then write the answer the way you would explain it to a friend. If you need a home for these articles, how to start a content website walks through setting one up.

What Real Data Showed Me on This Site

Here is why I keep saying “check it with real data”. When I connected this site to Search Console in September 2026, the page getting the most search impressions was an older article about Henry Ford’s success as an entrepreneur. The searches behind it were not from business owners at all. They were students pasting a quiz question whose answer is “his innovative production methods”.

No brainstorm, mine or a chatbot’s, would have guessed that audience. So I changed that page’s search title and description to give the short answer up front, and I am measuring whether more of those students click. That is the whole method in one example: let AI suggest, let real data decide, then write for the people who actually arrive.

What Mistakes Do Beginners Make With AI Keywords?

• Believing the numbers. If a chatbot gave you a search count, treat it as a guess until a real tool agrees.
• Chasing big phrases. A new site rarely wins the biggest phrases first. Start smaller.
• Writing for the tool, not the reader. Google says it wants helpful content made for people, not content made mainly to rank. It also says AI-written content is not banned, but using AI mainly to push rankings goes against its spam policies. Use AI to plan. Add your own examples, your own words, and real help.
• Making ten articles on one phrase. Pick one article per phrase. Two pages fighting for the same words just confuse everyone, including Google.
• Never looking again. Check Search Console after a month or two. Real data from your own site beats any guess.

What I Would Do This Week

Do not try to plan a year of content. Plan five articles.

1. Write your topic in one sentence.
2. Ask AI for 30 beginner questions, with no numbers.
3. Stretch your 10 strongest into longer phrases.
4. Check them in Google and Keyword Planner.
5. Pick five and write the first one.

That is a full day of work at most, and it gives you a clear list instead of a pile of guesses. If one of your ideas is about getting visitors, how to get traffic to a new website has ideas for what comes after the article is live.

Keyword research is not a talent. It is a habit. Ask AI for ideas, check them with real data, and write for the person on the other side of the screen. You can do that.

Questions People Ask

Can ChatGPT give me accurate search volumes?

No. A chatbot does not have live access to Google’s search data, so any number it gives is an estimate. Check it in Keyword Planner before you rely on it.

Is AI keyword research free?

It can be. Free chatbot plans handle the brainstorming, and Google Trends, Search Console and autocomplete cost nothing. Keyword Planner is free too, but Google asks you to finish a Google Ads account setup, including billing details, first.

Will Google penalise an article because I used AI to plan it?

Google’s guidance on AI content says using AI is not against its rules on its own. What it acts against is content made mainly to manipulate rankings, so plan with AI and write something that genuinely helps.

How many keywords should one article target?

One main phrase per article, plus the natural variations that come with it. If two of your phrases need the same answer, they belong in the same article.

When will Search Console show my own keywords?

Once Google starts showing your pages in results. On a brand-new site that can take a few weeks, so use the brainstorm-and-check routine above until your own data arrives.

You Might Also Like

• How to Get Traffic to a New Website
• How to Update Old Blog Posts for More Traffic
• Claude vs ChatGPT for Beginners: Which One Should You Try First?
• The Content-to-Income System
• How to Build a Niche Website With AI: The First 90 Days
• How to Use AI to Build an Online Business (Without the Nonsense)
• Why Pinterest Is a Traffic Source That Works for Camera-Shy Beginners

Get the free VivaMomentum Starter Kit