Choose blog SEO keywords from your own data
Adding candidates in an SEO tool does not decide the next article. A query with a high search volume is not always a problem you can answer, and sometimes fixing an existing article is enough.
Look first at words that already show up on your own site and in your ads. Collect Search Console, Google Ads search terms, and notes from inquiries and sales conversations. Check demand and competition in Keyword Planner last.
By the end of this article, you can use the CSVs you already have, plus free ChatGPT or Gemini, to separate rewrite candidates from new-article candidates and pick this round's one article with a reason.
The four data sources do different jobs
Do not collapse them into one score at the start. Separate what each source proves.
| Data | What it shows | What it does not show |
|---|---|---|
| Search Console | Which queries your pages were shown and clicked for | Search demand that never appeared |
| Google Ads search terms | The actual queries that triggered ads, and the results | Demand in areas where you are not advertising |
| Inquiries and sales conversations | Problems customers described in their own words | How many people search |
| Keyword Planner | Related terms, monthly searches, competition, and bid estimates | Whether you are trusted on that topic |
The Search Console API can combine query and page and return clicks, impressions, CTR, and average position. Google explains that the API does not guarantee every row and returns the top rows (Search Console API official reference). A missing number is not, by itself, zero demand.
The Google Ads search terms report is where you check the actual queries that showed an ad, and the results. Low-volume queries may not be shown individually, for privacy (Google Ads Help: About the search terms report).
Keyword Planner is for finding candidates and for estimates of monthly searches and cost (Google Ads Help: Use Keyword Planner). Competition there is how many advertisers there are. It is not how hard it is to rank in organic search.
Split Search Console into two lists first
From Search Console, export queries and pages for the past 28 to 90 days. A new site has few numbers, so if 28 days shows nothing, widen the range.
Split the table into these two.
Rewrite candidates for existing articles
An article for that query already exists. It is being shown, and it is not getting enough clicks.
- Average position is 11–30, and impressions continue
- It is inside the top 10, but it is not getting clicks
- The query intent and the answer in the title or body do not match
11–30 is not an official Google criterion. It is an operational band for finding candidates that are one step from page 1. If impressions are only 1, extend the period before you act on position.
For an article inside the top 10 that is not clicked, review the title and description before you add body text. Check whether the promise in the search result matches the answer the searcher wants.
New-article candidates
There is no article for that query. If only the wording differs and the search intent is the same, do not create a new one.
For example, 「Google広告 月次レポート 作り方」 ("how to make a Google Ads monthly report") and 「Google広告 レポート 手順」 ("Google Ads report steps") can be different keywords and still be answered by the same article. Group by what the searcher wants to finish doing, not by the string in the title.
Use ads and inquiries to see how close a query is to revenue
Next, add the Google Ads search terms report to the same table. Clicks are not the only thing to look at.
- Search terms that produced a conversion
- Search terms that spent money and did not produce a result
- Search terms whose intent does not match the ad or the landing page
A query with a conversion has, at least, led to an action through an ad. A query with a large cost cannot be called "high value" on that alone. The intent may have been off, and the clicks may have been wasted.
Line up inquiry and sales notes as one question per line. Even if search volume is small, a question customers ask repeatedly can become an article you reuse in sales or support.
The Google Ads numbers and the GA4 numbers do not match
I make the monthly report by hand every time
I do not know how to match ad spend against order data
At this stage, do not rephrase them into neat keywords. Keep the words the customer actually used.
Use Keyword Planner as the last check
After you have collected candidates from your own data, put them into Keyword Planner.
Check these three points.
- Can you confirm search demand?
- Is there a more natural phrasing, or a related term?
- From competition and bids, is this an area where advertisers are gathering?
You do not need to drop a candidate only because search volume is low. If it keeps coming up in inquiries and the problem is close to the product, it can also be used as sales material or a guide page, outside of search traffic.
How to narrow ad keywords from Keyword Planner numbers is in How to find search keywords with less competition. In this article, search volume is the last check on the theme.
Drop candidates on two conditions before you score them
After you join the data, do not build a total score immediately. Make these two conditions required.
- Can you answer it? You can show a basis for the article: product details, your own measurements, customer handling, and so on
- Is the action after reading decided? You can say, in one sentence, what the reader can do the next day
If a candidate fails either condition, do not write it, even if search volume is high. Then count how many of the next pieces of evidence overlap.
- It is showing in Search Console
- It has clicks or conversions in ads
- It comes up repeatedly in inquiries or sales conversations
- Search demand can be confirmed in Keyword Planner
Look first at candidates with two or more kinds of evidence, so one number does not pull the decision. If there is only one kind of evidence, and an existing article is in positions 11–30, you can still try a low-cost rewrite.
A prompt to hand to ChatGPT or Gemini
Attach the four data sources as CSV or text, and use the following as written.
From the attached data, choose the theme to cover this time on our blog.
The input is these four kinds.
- Search Console: query, page, clicks, impressions, average position
- Google Ads: search term, clicks, cost, conversions
- Keyword Planner: keyword, monthly searches, competition, bid
- Inquiries and sales conversations: questions customers actually used
Process them in this order.
1. Normalize only full-width and half-width forms, uppercase and lowercase, and repeated spaces
2. Group words with the same search intent. Write why you grouped them
3. If a matching existing article exists, mark it "rewrite". Otherwise mark it "new"
4. For each candidate, make a table of which input it appeared in and the original numbers
5. Check whether our product can answer it, and whether the action after reading can be said in one sentence
6. Output only the top 3, and briefly write why the others were not chosen
Do not guess search volume, CPC, conversions, or customer quotes that are not in the input.
Do not pick a winner by score alone. If an existing article can answer it, do not create a new article.
The point is to use AI for joining, sorting search intent, and checking duplicates, not for inventing candidate words. Whether demand exists is checked in the input data.
Choose only one article at a time
Saving 10 candidates does not add articles. Pick only the top one, and finish one of these.
- For an existing article, fix the title, the description, and the missing answer
- For a new article, separate its role from existing articles and write a draft you can publish
After publishing, review Search Console over the same length of period. Check whether impressions grew and whether the query intent widened, and connect that to the next rewrite. If position and clicks did not move, question the title, the search intent, and where the article lands before you add more articles.
What Shoryaku AI can do
Shoryaku AI can review data from Google Search Console, Google Ads, Google Analytics, and similar sources together, and make improvement moves with a basis. In chat, ask "From Search Console results, list the search keywords the next article should aim at," and it returns candidates from the data already connected. Once the theme is decided, ask in the same chat "Write a draft of an article on the theme of X," and the draft uses the business information you registered (how to ask is written up in another article).
It is not a feature that joins the four kinds of data described in this article and auto-publishes the article with the top score. You check the theme and the body before publishing. It is built so it does not write into ads or the blog automatically.
The 14-day free trial can be tried with no card at the start. How to hand the numbers you already have to AI is written up in How to analyze marketing data while talking with AI.