An idea for a blog post is only worth something when someone is already searching for that answer. Without demand behind it, it stays a title in a spreadsheet.
Ahrefs measured the cost of skipping this check on its own Content Explorer index, around 14 billion pages: 96.55% of pages get no clicks at all from Google, and another 1.94% get between one and ten clicks a month (study published in December 2023, global sample).
For a site that has been live for a few months, the most reliable source of ideas is not a title generator. It is the Performance report in Google Search Console, which lists the searches your site already appears for without anyone having written a line for them. Below are the four filters we use to read it, real data from two sites we work on and the eleven tools to move on to when that data runs out.
Why idea generators return lists you can't use
A generator takes a word and returns title variations. HubSpot's Blog Ideas Generator has done this for years; ChatGPT and Perplexity do it faster and with more imagination. None of the three knows how many people search for that thing, or whether your site has any chance of answering better than whoever is already on page one.
You can see where this leads on Google UK. For "blog ideas" (about 210 searches a month) the first page opens with an AI Overview that lists profitable niches and links to HubSpot's generator, followed by lists of 100 or 120 ideas from Wix, Adobe and lifestyle bloggers (DataForSEO, Google UK, September 2026). Plenty of titles, no demand data.
For an idea to become an article you need three conditions at once:
- a question people actually ask, with a search volume or at least impressions to prove it;
- something of your own to put into it: a number, a case, an objection you hear when quoting;
- a link to what you sell, otherwise the traffic arrives and produces nothing.
The third condition is the one that removes the most titles. A packaging machinery manufacturer can write about industrial sustainability and collect a few visits, but quote requests come from people searching for "fresh pasta packaging machine". If the list of ideas ignores this, you end up with a full calendar and an empty funnel: this is the starting point of a content marketing plan built on numbers.
The four filters in the Search Console Performance report
The Performance report shows four figures for each search: impressions, clicks, CTR and average position. Google Search Central's documentation presents it as the tool for seeing how much traffic comes from Google Search, broken down by query, page and country. Almost everyone opens it to see whether traffic is going up or down. It also tells you what to write.
1. High impressions, low clicks, good position
The site appears often and high up, and almost nobody clicks. In most cases this means the page Google shows is not the right one for that question: a homepage or a product page is ranking where an article is needed.
2. Average position between 8 and 20
These are searches one step away from page one, reached by a page that was not built for them. A targeted article on that query usually gains more positions than any technical work on the site.
3. Queries written as questions
Filter the queries containing how, what, how much, why, which. They are titles already written by users, and they are also the form AI assistants pick up when they cite a source.
4. Queries no page covers
Cross the list of queries with the list of pages. Every search that sends impressions to the homepage or to a generic category is a gap in the editorial plan, and therefore an idea.
Here is how it reads in practice. The table below comes from the Search Console of an Italian LED screen manufacturer, limited to searches for which the site gets impressions without having a dedicated page. The queries are in Italian; the English meaning is in brackets.
| Query | Impressions | Clicks | CTR | Average position | Filter | Resulting idea |
|---|---|---|---|---|---|---|
| ledwall significato (LED wall meaning) | 3,511 | 32 | 0.91% | 4.67 | 1 | A page that defines the term |
| ledwall cos'è (what is a LED wall) | 3,435 | 25 | 0.73% | 4.93 | 1 | The same page as the row above |
| schermi led (LED screens) | 2,840 | 11 | 0.39% | 14.77 | 2 | Category guide |
| led wall outdoor | 2,134 | 21 | 0.98% | 14.43 | 2 | Article on outdoor installation |
| ledwall prezzi (LED wall prices) | 1,648 | 33 | 2.00% | 8.85 | 2 | Pricing page with real ranges |
| ledwall costo (LED wall cost) | 795 | 39 | 4.91% | 10.68 | 2 | Same pricing page |
| software per ledwall (LED wall software) | 551 | 11 | 2.00% | 2.52 | 4 | Article on management software |
| ledwall farmacia (LED wall for pharmacies) | 501 | 10 | 2.00% | 13.78 | 4 | Vertical use case |
| ledwall bordo campo (pitchside LED wall) | 283 | 10 | 3.53% | 11.07 | 4 | Vertical use case |
Two rows deserve attention. "LED wall meaning" and "what is a LED wall" are the same question written two ways: 6,946 impressions combined, 57 clicks, average position around fifth. The site is visible and people don't click, because for a plain definition the box at the top of the page now answers directly.
The last two rows say the opposite. "LED wall for pharmacies" and "pitchside LED wall" have small numbers but describe a specific installation, meaning someone with a budget and a problem. In terms of quote requests, 501 impressions at position 13.78 are worth more than 3,511 impressions on a definition, which is why long-tail keywords deserve a place in the calendar even when tools call them irrelevant.
Seven searches, one article
The second case is an Italian solar farm developer. Its Performance report contains a group of almost identical queries, all with zero or one click.
| Query | Impressions | Clicks | Average position |
|---|---|---|---|
| vendo terreno per fotovoltaico (selling land for solar) | 96 | 1 | 7.61 |
| affitto terreno fotovoltaico (solar land lease) | 89 | 0 | 18.84 |
| affitto terreni per fotovoltaico (leasing land for solar) | 55 | 0 | 11.44 |
| affittare terreno per fotovoltaico (lease land for solar) | 44 | 0 | 15.30 |
| vendita terreno per fotovoltaico (sale of land for solar) | 31 | 1 | 12.42 |
| affitto terreno per pannelli solari (land lease for solar panels) | 20 | 0 | 10.90 |
| agrivoltaico affitto terreno (agrivoltaic land lease) | 20 | 0 | 8.15 |
These are seven ways of asking the same thing: on what terms a landowner can sell or lease land for a solar farm. In total 355 impressions and two clicks. You don't need seven articles, you need one, written for the landowner and not for the investor. The rule is simple: if two phrasings return the same first page on Google, they are the same question, and search intent is read on the SERP, not in the word count.
The same list contains a row of a different kind: a question written out in full, in English, in the form "who is the ceo of", with 43 impressions on an Italian-language site. That is not how people type into Google, it is how they write to an assistant. Rows like that are becoming a second list of ideas, and they should be read separately.
When Search Console doesn't have enough data yet
A new site, or one with a few hundred impressions a month, doesn't produce a readable list. In that case the ideas are in four places, and none of them is a tool.
Quote request emails and sales calls. Many copywriters build their whole method on this: collecting the questions clients ask again and again and turning them into articles. It is the only source that hands you the question already phrased by someone with the money to buy.
After-sales support. Repeated tickets are missing manuals. A company that gets the same request twenty times a year has an article already written, and saves hours on the phone too.
Reddit and Quora. Use them for the words, not the numbers. The way a user describes a problem in a thread is often closer to the real query than any tool output.
The "People also ask" boxes. AlsoAsked maps them as a tree and says it takes the data directly from Google. They show the structure of related questions, useful for deciding what goes in the subheadings and what to leave for a second article.
Where the number you're reading really comes from
Tools for finding ideas are not interchangeable, because they look at different sources. Before trusting a figure, it pays to know where it was taken from.
| Tool | Where the data comes from | Good at answering | What it doesn't tell you |
|---|---|---|---|
| Google Search Console | Your site, measured by Google | What you already appear for without trying | Very rare queries, which Google filters out |
| Google Trends | Sample of Google searches, relative index from 0 to 100 | Whether a topic is rising or falling | How many people in absolute terms |
| Keyword Planner | Google Ads, rounded and grouped volumes | The order of magnitude of demand | Differences between close variants |
| Ahrefs, Semrush, Moz | Keyword databases plus clickstream panels | What competitors rank for | New queries and very local ones |
| AnswerThePublic | Questions and prepositions built around a term | How many ways the same thing is asked | Whether anyone actually searches for them |
| AlsoAsked | "People also ask" boxes | How Google connects questions to each other | Search volume |
| BuzzSumo | Social shares and engagement | What circulates, not what is searched | Demand on Google |
| HubSpot Blog Ideas Generator, ChatGPT, Perplexity | Language generation | Title variations and possible angles | Whether the demand exists |
The column that matters is the second one. Three entries in this list measure nothing: they produce text. They are fine for rewording a title, not for deciding whether to write it.
The 11 tools and the question to ask each one
- Google Search Console. Free. Question: which searches do I already appear for without a dedicated page?
- Google Trends. Free. Question: is this topic seasonal, growing, or have I caught the tail end of it?
- Google Ads Keyword Planner. Free with an active account. Question: roughly how big is the demand for this group of terms?
- Ahrefs. Paid, with a reduced version for verified site owners. Question: which competitor pages bring in traffic, and for what?
- Semrush. Paid. The Topic Research tool groups subtopics and headlines around a theme. Question: which subtopics keep coming up among the pages that rank?
- Moz Keyword Explorer. A few free queries, then paid. Question: does this query exist in the UK with usable volumes, or have I only seen it in US data?
- Ubersuggest. Question: which variants of the keyword have less competition?
- AnswerThePublic. Question: in how many different ways is this question phrased?
- AlsoAsked. Question: which questions does Google consider related to this one?
- BuzzSumo. Paid. Question: is this topic discussed outside search engines?
- Reddit and Quora. Free. Question: what words do people who really have the problem use to describe it?
Title generators are left out of this list on purpose. Not because they are useless, but because they belong at the end, once the idea has been validated, to choose how to phrase the headline.
Four checks before you open the document
An idea goes to writing only if it passes all four.
- Does the demand exist? Impressions in Search Console or volume in at least two different databases. How to cross-check them is covered in the guide to keyword research starting from the data you already have.
- What format does the first page in your market want? If the top results are service pages, an article won't get in. If they are guides, a product page won't get in.
- Do I have something the others don't? A figure, a measurement, a case with numbers. Without it the article is the tenth copy of the same page.
- Can the people who arrive from it buy? Not every idea has to bring in clients, but if none does, the blog is a cost.
The second check is the one most often skipped, and it produces the most expensive mistake: articles written for a query where Google shows only commercial pages. It is worth checking even when the SERP shows a featured snippet, because in that case the short answer is already taken.
Ideas that return less today than three years ago
Definition questions are the first that come to mind and the first to have lost value. Pew Research Center tracked the browsing of 900 US adults in March 2025, across 68,879 Google searches, 12,593 of which showed an AI summary: users who saw the summary clicked a result on 8% of visits, those who didn't on 15%. A US sample.
Ahrefs reached a similar conclusion with a different method: 300,000 keywords compared between March 2024 and March 2025, US market, and a 34.5% drop in clicks for position one when an AI Overview appears.
The equivalent from our own data, in the first table of this article, is the "what is a LED wall" row: average position 4.93 and a CTR of 0.73%. A site visible in fifth position for a definition gets less than one click for every hundred times it is shown. That doesn't mean you should stop writing definitions; it means you shouldn't expect them to bring traffic. They are there to be cited, and the place to check that has changed. Since 31 August 2026 Search Console shows every site reports on AI features, with impressions for AI Overviews and AI Mode split by page, country and device, but without clicks. The wider picture is in our analysis of the impact of AI Overviews on SEO and the one on zero-click searches.
From the list to the calendar
A list of forty ideas is not an editorial plan. It becomes one when every entry has a priority and a date, and when part of the space is reserved for updates rather than new pieces.
Updating matters more than it used to because producing text keeps getting cheaper for everyone. According to the ONS article Artificial intelligence in UK businesses, published on 20 July 2026 with data from the Business Insights and Conditions Survey, the share of UK businesses with ten or more employees using AI rose from around 12% in late 2023 to around 35% in June 2026, and among businesses with 250 or more employees it reaches 49%. The volume of published content is rising, so the idea alone no longer sets anything apart: what does is the evidence you put into it, and the published content you keep up to date.
A pace that works for a small company: two new articles and one update a month, with a quarterly check on pages losing clicks while impressions stay flat. What to check and how often is in the SEO checklist, and on how much to write for each piece we collected the data in the article on ideal blog post length.
An example of where this way of choosing topics leads: in the SEO for manufacturing case, work on a cluster of commercial queries took macropix.it to a 25.55% share of voice on the LED wall cluster in July 2026, ahead of amazon.it at 18.08%. It didn't come from an idea generator; it came from a list of queries read one by one. If you want to get there with a structure already in place, this is the work we do in our SEO services.
FAQs on blog post ideas
Without Search Console data you start from the questions that arrive before the search: quote request emails, sales calls, support tickets. Then you check them on Reddit, Quora and in the People also ask boxes, and finally you make sure there is search volume in your market in at least two different databases.
It helps with rewording and finding different angles, not with deciding what to write. A language model generates plausible titles but doesn't measure demand: it doesn't know how many people search for the topic or in what form. The idea always needs checking against a source that measures, such as Search Console or a keyword database.
For a small company, two new articles a month plus one update is a sustainable pace, so you need about 24 validated ideas and around ten published pages to revise. A longer list than that usually goes stale before it is used.
If an existing page gets steady impressions and falling clicks, an update returns more than a new piece, because it starts from a position already gained. A new article makes sense when no page covers the query or when the page Google shows answers a different question.
To get started, yes. Search Console, Google Trends, Keyword Planner and the related questions boxes cover the part that matters, which is knowing whether the demand exists. Paid tools are mainly for seeing competitor data and for working on many sites at once.
The most valuable list is the one of rejected ideas
Almost every company keeps a list of topics to write about. Almost none keeps the list of rejected topics with the reason next to each one, and that wastes hours in every meeting, because the same idea comes back every three months and has to be assessed from scratch each time.
Two columns are enough: the idea and the reason it was dropped, for example no volume, SERP taken by service pages, no data of our own to bring. After a year that sheet becomes the most useful part of the editorial plan, because it describes the real boundaries of the market in which the company has something to say. And every so often a row should be reviewed: ideas rejected for lack of demand sometimes build that demand on their own.
