If you know your product best, I think you should write its SEO content. You know what it helps people do and where it falls short. When someone asks why they should choose it over another tool, you have something to say.

Founders often hire writers or an SEO agency instead because they don’t have time to write. With AI, you can do much more of that work yourself: talk through an idea, get help with research, and edit a draft without writing every sentence from scratch. Marketers call this founder-led SEO, and most guides to it are written by agencies selling the service.

An SEO team may know which keywords to target without understanding the product. If the work starts and ends with search data, you can end up with articles that cover the keywords but barely explain why anyone would use what you’ve built.

For a comparison article, you need to explain when switching to your product makes sense and when someone should keep what they already use. Search volume won’t answer that. You need to understand both products and the work people do with them.

I’ve been using AI to write StashBase’s SEO content, and I’m happy with the results. I discuss my ideas with the agent, check the outline, then work through the draft. I still need to check the facts and edit what I wouldn’t say.

Start with a question you can help answer

Choose a question someone might ask while deciding whether to use your product. Look at support conversations and public discussions for the words people use. Search data can help you decide which questions to investigate first.

For StashBase, one of those questions is what to use if you want something like Obsidian, with AI built in. That gives the article a reader and a decision to help with.

Someone asking that may already have notes they want to keep. They may pay for an AI subscription and want to use it with those notes. Before recommending anything, we need to find out whether their current setup can do the job and what changing tools would involve.

You can ask AI to suggest related questions. Treat those as ideas to check against actual conversations. An agent can invent a plausible reader just as easily as it can invent an answer.

Give AI your answer before asking for a draft

Write a few rough notes about what you would tell that reader. Include your recommendation and the reasons behind it. If you’re unsure, say what you need to check.

For the Obsidian example, a useful starting point is: “If someone is happy with their notes app, I’d first check whether they can add the agent they want. To recommend switching, we need to explain what gets easier and what they would lose.”

That gives the research a purpose. The agent can check the available integrations, costs, and limits before either of you settles on a recommendation.

If you find it easier to talk than write, use a short prompt:

I want to answer [reader’s question]. Ask me what I’d recommend and why, one question at a time. Help me identify what we need to verify before drafting.

Keep the answers in a brief beside the draft. Include who the reader is and what the article should help them decide. Correct the brief when the discussion changes your mind.

Check the answer, then work through the draft

Verify product claims before asking AI to turn the brief into an outline. Read current documentation and try the workflow when the answer depends on how it behaves. Save the source links and record what you actually tested.

Knowing your own product doesn’t tell you what a competitor supports today. A feature you think is missing may already exist through an integration. Give the other product the same attention you’d want someone to give yours.

Read the outline for the answer it promises. Does it address the reader’s conditions? Is there a section that exists only because articles about this topic usually have one? Cut that before it becomes several paragraphs.

Then ask for a draft. Read it as if you were answering the question yourself. Replace a vague benefit with what someone can do. If a paragraph says a tool “improves productivity,” ask which part of the work gets easier and how.

Check factual claims against the sources. Keep links beside the claims they support. For instructions, follow the steps yourself and include the relevant plan or setup conditions. Don’t let the agent describe something as tested if neither of you tested it.

For wording, give it an article you like and explain what to follow. I prefer short explanations and concrete examples. Our guide to writing with AI without losing your voice goes into how to use your own edits as writing preferences.

Keep the material for your next article

Keep your product notes and finished articles together so you can reuse them. Save source links with enough context to know what they supported. A pricing claim checked last month may need checking again; a writing sample can still show how you like to explain things.

This is where I use our app, StashBase. I keep the writing material in a local folder and work with an agent beside the documents. I can open a source while reviewing the draft and see the agent’s file edits. The agent setup guide explains how to connect one.

StashBase also prepares text from PDFs and DOCX files so you can search those sources alongside your notes. Search by meaning helps find a relevant passage when you don’t remember its exact wording. As the material grows, you can ask the agent to build a Wiki that organizes the topics and links back to sources.

For the next article, ask the agent to find related material before you start. It may find a test you can reuse or an existing article that already answers the question. Sometimes updating that article is enough.

Publish, then see whether it helps

Before publishing, check that the title matches the question and the opening answers it. Link to related articles where they save the reader another explanation. When you recommend your own product, say it’s yours and explain why it fits.

Once the article is published, use Google Search Console’s Performance report to see which queries show your page and bring clicks. Read those queries alongside the article. They may reveal a question you haven’t answered clearly.

Use your site analytics to check whether readers go on to explore the product or sign up. An impression, a click, and a signup tell you different things. Keep those results separate when deciding what to write next.

My satisfaction with an article doesn’t establish that it brings customers. I still need to look at what happens after publication. The same applies when AI makes it easier to publish more.

When to hire an SEO writer or agency

Writing your own content still takes time. If you can’t keep up, an editor or an SEO specialist can help with research, writing, and finding problems you’ve missed.

Give them access to the product and your reasoning. Let them try the workflows and ask you questions. If someone on your team understands the product better than you do, involve them in the article.

A brief with your recommendation, supporting evidence, and known limits gives another writer something useful to work from. As you hand over more of the writing, keep reviewing the claims and examples until you’re confident they understand the product well enough to make those calls.