If you run a small SaaS company, you probably have a product worth explaining and no clear idea of what to publish, while competitors seem to have an endless supply of topics. This guide covers how to find blog post ideas in a way you can repeat: you start with real buyer questions, check them against search data and against what already exists, and end up with a short list of posts you can defend before anyone writes a word. It works for a niche or unglamorous industry too, because the questions your buyers ask are the subject matter, whether or not the industry is exciting.
This is a process for finding and checking topics. It isn't a plan for your whole blog, and it isn't a list of headline prompts you can swap in and out.
Step 1: Define the buyer and the question the post must answer
Pick one buyer, not "everyone interested in our industry." Write down their role, the task or decision they're working on, what they probably already know, and what would make an answer useful to them. Put that in one line at the top of your working list, so every idea you collect afterward gets held up against it.
It helps to separate a prospect who is evaluating a tool from an existing user who is troubleshooting it. A support issue can make very good documentation without being a good prospect-facing blog post, and it's worth deciding early which one you're after.
You're done with this step when someone else could look at any candidate on your list and tell who it's for and what that person could do after reading it. The usual mistake is starting with a fashionable keyword or a competitor's headline and inventing the audience afterward. Google's guidance on helpful content asks whether a site has an intended audience, offers original value, and leaves readers able to reach their goal, so it's a good check on your one-line buyer description.
If you keep your personas and product details in one place, this step gets easier. In DeepSmith, Deep IQ stores your buyer personas and product context, so the buyer you chose stays available when you start collecting ideas. It still needs your input about who your customers are, and it won't discover them for you.
Step 2: Collect the questions customers already ask
This is where blog topic ideation gets its best material. Go through recent sales calls and onboarding conversations, support tickets, customer-success notes, demo objections, prospect emails, and product feedback. Ask anyone who talks to customers for the actual question and the situation it came up in. Record the wording, who asked, where it appeared, and what answer was missing.
Then group near-duplicates by the decision underneath them. Three people might phrase something three ways, and that's still one post. Remove names and confidential details before anything goes into a shared backlog.
If conversations are sparse, ask a recent customer or prospect what they tried to find out during a real decision, what confused them, and which explanation helped. Qualitative feedback like this can show you the why behind a pattern that analytics can't explain. HubSpot describes teams collecting customer questions in a shared channel and finding recurring sales-call pain points that turn into content. One example it documents is a leave-management software team that noticed prospects wanted the product but needed leadership buy-in, so they wrote about how HR could make the case to a CFO.
You're done when each shortlisted question has either a traceable conversation or clearly marked outside evidence, along with a proposed reader and the need behind it.
Common mistake: treating one loud request as proof that the whole market asks the same thing. One conversation is a lead to look into. It also helps to keep the buyer's exact words, because they often become the best phrasing for search later, and to avoid turning a private support exchange straight into a public case study.
Here is a made-up example to show what a good candidate looks like. Say you sell software for industrial maintenance teams. "How do I tell a sensor alert from a real maintenance issue?" is a much better candidate question than "Exciting trends in industrial software." It still needs to be checked against real customers and real search results, so treat it as an illustration and not a documented quote.
Step 3: Check what your site already attracts in Google Search
Now look at what people already find you for. In Google Search Console, open the Search results Performance report and pick a date range and search type. The default view covers the past three months, and you can change it. Open the Queries tab and sort by impressions or clicks, then note any relevant terms, especially ones phrased like a question. Open the Pages tab to see whether an existing page already covers a candidate. You can also click a query and then look at Pages to see which of your URLs showed up for it.
Clicks are visits from Google Search results, impressions are appearances, and click-through rate is clicks divided by impressions. Average position is the average position of your topmost result, so it isn't a score for how good an idea is. Google suggests paying more attention to trends in impressions and clicks than to position alone. A low click-through rate can point to a mismatch, but look at the query, the page, and the title first, because it doesn't automatically mean you need a new article.
What you want here is a question that already gets impressions but has a thin answer, and any page that should be improved instead of duplicated. You're done when every promising query is paired with its existing page, or marked as an open coverage question you still need to look into. Keep "improve existing page" separate from "create new post" on your list.
Common mistake: treating Search Console as a full picture of buyer demand. It shows how your own site performs in Google Search, not everything people search in your category. Some queries are withheld for privacy, and a new site may have very little data at all. If yours is new, skip ahead and come back to this step later.
Step 4: Compare competitor coverage with what your buyer needs
Pick the real alternatives your buyer considers, and read their blog posts, documentation, and product explanation pages for questions that keep showing up. For each candidate, note whether you have no answer, a thin answer, or a good one. Read the pages themselves, because counting titles tells you very little. Then ask what you could add that they haven't, such as a use case, a constraint, a demonstration, or something you learned firsthand.
You're done when each competitor-gap idea names the buyer question, the answers that already exist, the specific value that's missing, and whether an existing page of yours could be strengthened instead. A competitor publishing a new page is a reason to investigate, not a reason to publish.
Common mistake: copying the top-ranking article's structure and calling that a gap. Google asks whether a page offers substantial added value rather than a rewrite of other sources, and that's a fair test for your own ideas too.
This is a step where DeepSmith does real work for you. The Content Map turns your site and your competitors' sites into one map of topics, so coverage is something you can measure. Coverage gaps show topics where you publish but a competitor publishes more, and untapped topics show where a competitor writes and you have nothing. Competitor sitemaps are checked daily for new pages, so you don't need to go looking for them. It finds candidates, and you still have to decide whether a topic fits your buyer and whether you can say something useful about it.
Step 5: Expand the survivors into search language
By this point you should have a handful of questions that survived. Start your keyword research from those questions, not from a random seed term. Try the customer's own wording, plain-language alternatives, and terms you noticed on relevant pages. In Google Ads Keyword Planner, the Discover new keywords option returns candidate expressions, and you can refine them by keyword text, average monthly searches, categories, location, and language. Set the location and language to the market your buyer is actually in. Google notes that basic idea-generation features need a completed Ads account setup, including billing information, so plan for that.
Google Trends works well as a second signal. It compares relative interest across terms, places, or periods, and its results are normalized from 0 to 100, so it isn't a count of searches. Low-volume terms can show 0 even when some people do search them, and small spikes can be noise. Neither result should cancel out a niche question you have good evidence for.
You're done when each candidate has a main buyer question, a few search phrasings, a target location, and a plain note if the search data is weak or missing. There's no agreed minimum monthly search number that makes a topic worth writing, so don't invent one.
Common mistake: picking the biggest number without checking who is searching it, or reading the competition label in Keyword Planner as a promise about how hard it will be to rank organically. That label is built for advertising.
Step 6: Inspect what searchers actually receive
Search your best wording and read what comes back. Notice the types of results and the leading answers, and work out whether the person wants instructions, a definition, troubleshooting, a product comparison, or to buy something. Then write down what those answers cover and what they leave out. If the topic matters in AI search, try a similar conversational question in an AI tool and look at the answer and the sources it links, as one more thing to learn from.
Then adjust the idea. Narrow it so it answers one coherent need, or drop it if a blog guide is the wrong format. You're done when the format you picked matches what the searcher appears to want, you can say in one sentence what's missing from today's answers, and your team can bring real product knowledge, examples, or firsthand experience to fill it.
Common mistake: assuming that AI answers give you a guaranteed way to get cited. Google says its established SEO practices still apply to its AI features, and meeting the eligibility requirements doesn't guarantee inclusion. An AI answer is something to observe and follow up on. It doesn't prove that readers want the topic, and it doesn't predict any result.
Pro tip: when a niche phrase looks tiny in a keyword tool, check whether a small number of qualified buyers keep asking the same underlying question. Look at their conversations together with the search evidence you can observe, and don't quietly let one stand in for the other. With blog post ideas for niche industry topics, a few qualified readers who really need the answer are often worth more than a large number of casual ones.
Step 7: Validate the answer you can uniquely give
Before you keep an idea, ask someone who knows the product or the customers whether you can answer this exact question accurately. Which steps, limits, screenshots, calculations, or firsthand observations can you actually back up? Would the answer still help a reader who never buys from you? If search results already satisfy the need and you can't add anything meaningful, discard the idea or narrow it. If the answer would need customer data, get permission first, or use an example that is clearly labeled as hypothetical.
You're done when each idea you're keeping has a reader, a question, an intended outcome, a format, something that sets your answer apart, and a named expert or other source behind its claims.
Common mistake: handing a nuanced product topic to a generic draft with no source of truth behind it. Whatever tools you use, someone who knows the subject still needs to check the facts before it goes out.
Step 8: Choose the next few ideas and keep the evidence
This last step is how to choose blog topics without going back to brainstorming. Make one row in your backlog for each idea that's still standing. A row can hold the buyer and their decision, the original question and where you heard it, the related search terms and location, your existing page if there is one, the gap you found in competitor or search results, your proposed angle, the evidence or expert you need, and a decision to create, update, defer, or reject.
Order the rows by how close each is to a real buyer need, how strong the evidence is, how well you can add value, and how much effort it takes to answer accurately. Those are editorial judgment calls, and they aren't a scoring formula with proven weights, so adjust them to fit your team. Come back to the list whenever new questions arrive.
You're done when you could hand one row to a writer and explain why this post, for this reader, is worth making now, without asking for another round of broad brainstorming. A list of titles with no buyer question or evidence behind them is the usual failure, and so is having several posts aimed at the same need when one improved page would do.
DeepSmith can hold this list for you. Opportunity Agents read your AI visibility data or your Content Map and return ideas with the data point that justifies each one attached, and New Ideas is the single backlog for ideas from agents, manual entry, or import. That keeps your candidates and their reasons in one spot. Confirming that the buyer's question is real is still your job.
What to do next
Pick one validated question this week. Decide whether it should update a page you already have or become a new post, and save the evidence with the idea so a writer can pick it up later. Once you have a content ideation framework you trust, the next round of ideas comes from the same steps with newer questions, newer search data, and whatever your competitors have published since.
If you'd like help spotting coverage gaps and keeping write-ready ideas in one place, you can start a DeepSmith free trial and try it on your own site.
