How to find out what your customers actually search for: Wordstat and Google Trends

To find out what people actually type into search in your field you need two free tools and about fifteen minutes. Yandex Wordstat (a Yandex tool that shows how many times a phrase was searched in a month) gives you numbers. Google Trends (a Google tool that shows how interest moves over time) gives you the wordings that are growing right now.

The order goes like this. Set your region in Wordstat first, because by default it counts the whole world and the numbers come out inflated. Type 3-5 short words from your topic one by one, not questions. Copy the column with the numbers until you have 20-40 lines in the format "query, this many searches per month". Do not clean anything up: junk inside the list is normal. Then hand the list to an AI and ask it to sort the lines by the intent behind them, into four groups: understand, fix, choose, buy. The first three become articles. The "buy" group never becomes an article, that is what a product page with a price is for.

Below is every step in detail, including the settings people usually skip.

How is Wordstat different from Google Trends?

Wordstat answers the question "how many". It shows how many times over the last 30 days people searched Yandex for phrases containing your word. Its trend data goes back to 2018, and results from any tab can be exported with the "Download" button as CSV or XLSX (service description in the Yandex Direct knowledge base, checked 10 August 2026).

Google Trends answers the question "which way is this moving". It gives no absolute search volume at all, only relative interest. What it does give you is the "Related queries" block with a Top / Rising switch. Rising shows growth as a percentage against the previous period, and instead of a percentage you may see the word Breakout, which means the term grew by more than 5000% (Google Trends help, checked 10 August 2026).

Practical takeaway: take numbers from Wordstat, take fresh wordings from Trends. Mark the Trends lines in your list with the word "Trends" instead of a number, otherwise you will later mistake relative growth for search volume.

One honest note if you sell outside Russia and the CIS: Wordstat only counts Yandex searches, so the numbers part will not apply to you. Everything below about intent, sorting and product pages works regardless of which search engine your customers use.

How do you set the region so the numbers are not inflated?

Under the search box in Wordstat there is a region selector, and by default it is not narrowed down, the whole world is counted. If you work across Russia, set Russia. If you deliver flowers in one city, set that city.

The difference is not cosmetic. A query with a few thousand searches "everywhere" may show twenty in your city. The reverse also happens: a narrow topic that disappears in global data turns out to be perfectly workable in your own region. Set the region before you start copying, rebuilding the list later costs more.

Wordstat needs a Yandex account, free, the same one you use for mail. Google Trends opens without any sign-in (checked 10 August 2026).

Why type a short word instead of a question?

Wordstat looks for phrases that contain what you typed. Type "claude" and you get hundreds of variants: "claude ai", "claude free", "claude or chatgpt". Type "how do I work with the claude ai model" and you get one line with a tiny number and a nearly empty column next to it.

The rule is simple: one or two words, the most general name of what you sell, plus the synonyms your customers use, not the ones you use. Run 3-5 such words one after another. Leave your company name out of this list: people who search for it already know you, and you are collecting the ones who do not.

How do you gather 20-40 lines without losing an evening?

The goal of this step is a note or file where every line looks like "query, this many searches per month". Fewer than twenty lines is not worth it: there are four intent groups ahead, and after cleaning you would be left with a couple of queries each. More than forty on the first pass is pointless too, you will not work through a quarter's worth of topics anyway.

Junk inside the list is fine. Other cities, adjacent niches, random words that happen to share your term. Do not clean at this step at all, cleaning is a separate step and it is not done by hand.

Two ways that work:

• Copy it yourself. Select the column with the mouse, paste it into a note, repeat for every word. About fifteen minutes per topic. There is a "Download" button too, but with five runs, five files are usually less convenient than one note.
• Install the Claude extension for Chrome. Then the AI sees the open Wordstat page itself, and "collect the queries and numbers from the left column into one list" is enough.

One thing you must not do: ask an AI for search volume. It will produce a plausible number that does not exist, you will put it into your plan and spend a week on an article for a query nobody types. Search volume is counted by Wordstat, and it gets into your list by hand.

How do you sort a raw list by intent?

Intent is the reason a person went to search in the first place. The same word in different wordings marks a different stage, and each stage needs different content.

Four groups:

• Understand: "what is", "how does it work", "how to use".
• Fix: "not working", "error", "what to do if".
• Choose: "or", "which is better", "alternatives", "comparison".
• Buy: "price", "cost", "buy", "order", "services".

Sorting forty lines by hand is tedious, and it is exactly the work an AI does in a minute. Give it the list in one block and ask it to go in order: drop the lines that are not about your business and the ones too broad to write a single text about, give a one-line reason for every line dropped, sort the rest into the four groups, find duplicates (one meaning in different words), and for every group except "buy" suggest headlines for future articles.

A headline should read as the literal question a person asks, in the words from their query. "How much does outsourced bookkeeping cost for a sole trader" is a headline. "Everything about choosing an accountant" is a magazine section, and nobody searches for it.

When the table comes back, check three things. No articles are planned under the "buy" group. Every dropped line has a stated reason instead of quietly disappearing. Debatable lines you moved yourself: you know your niche better than any model does.

Why does a "buy" query need a product page and not an article?

Someone typing "buy X price" has already decided. They need the price, the terms and a button. Landing on an article called "10 criteria for choosing X", they go back to search and end up with whoever answers immediately.

So the "buy" group is served by a product or service page, and the words from the query belong on that page: in the heading, in the first paragraph, next to the price. If that page does not exist yet, write down what should be on it and build it before you start writing articles.

Articles are for people still figuring things out and comparing options. A product page is for people who have chosen and are looking for where to pay. Mix the two up and both run empty: the article does not sell, and the page brings in no new people.

What if a group is empty after cleaning?

That is not a broken prompt, it is an honest answer: for that intent you have nothing to write yet. Go back to Wordstat and gather words specifically for the empty group. Nothing under "fix", search with "error", "not working", "broken". Nothing under "choose", add "or", "alternatives", "which is better" to your word. Add the new lines to the same chat and ask for the table to be rebuilt.

The opposite happens too: the AI drops a query that is obviously yours. Say so plainly, "put that line back, this is one of my main services", and explain why in one sentence. It will not argue.

What do you do once the table is ready?

Take the top five lines by priority and put dates on them: that is a month of work, one article per week. Without dates the list stays a tidy file nobody opens again.

A month later run the same 3-5 words through Wordstat and compare the numbers. Seasonality and new wordings show up that way, not by gut feeling.

A detailed walkthrough with ready-made prompts is in the "Echo" block of the Academy, three lessons are open for free: academy.agineai.com/block/b12-eho.