Oops! Something went wrong while submitting the form.
Every guide tells you to enter a competitor's domain. None of them tell you how to know who your SEO competitors actually are. Sai reads your own content, extracts the terms you already rank for conceptually, and gives you the seed list that determines who you should be analyzing.
The recording is a real session. The sheet on the right is what it produced.
Sai opens each profile, pulls the signal, and writes the row, live, in a real browser.
Eight columns, sorted by score, with a source link behind every claim.
Your website URL, and any content you want included — a blog, docs, or landing pages.
A grouped list of seed keywords drawn from your actual content, sorted by theme and intent, with the likely competitors for each cluster.
Around 10 minutes.
Re-run each quarter, or after any significant content push, to catch the themes your newer pages introduced.
It's the practice of examining who ranks for the searches you want, and working out what they're doing that you aren't.
The standard process is well documented. You identify competing domains, run them through a tool like Ahrefs or Semrush, pull their ranking keywords, find the gaps between their coverage and yours, and prioritize what to write. Every major guide describes some version of this and the process genuinely works.
It has one unexamined step, though, and it's the first one.
Almost every guide skips this. Step one is "enter your competitor's domain," which quietly assumes you already know whose domain to enter.
For most companies that assumption is wrong, because SEO competitors and business competitors are different sets of people. The company you lose deals to might publish nothing at all. Meanwhile the pages actually outranking you might belong to a media site, a forum thread, a directory, or a company selling something unrelated whose content happens to cover the same problems your buyers search for.
Feed your sales-competitor list into a keyword tool and you'll get a beautifully detailed analysis of the wrong websites. Everything downstream inherits that error, and nothing in the tool will tell you it happened — the report looks equally professional either way.
Your SEO competitors are defined by the searches you want to win. So you can't identify them until you know what those searches are. Which makes keyword discovery the actual first step, and competitor analysis the second.
Seed keywords are the short, foundational terms that describe what you do. They're the input to everything else — you expand them into long-tail variations, you use them to find who ranks, and you use them to structure your site.
Get them right and the rest of the process compounds. Get them wrong and you'll do excellent research in the wrong territory.
The usual way people generate them is by brainstorming in a document, which produces a list of terms describing how the company thinks about itself. That's the failure mode. Internal vocabulary and search vocabulary diverge constantly — you call it a "workflow orchestration layer," your buyers search for "how to automate reporting." A seed list built from your positioning deck will be full of words no one types.
The more reliable source is your own published content, because content written for readers tends to use the language readers actually use. It's evidence rather than aspiration.
It reads your site the way a person would, then extracts terms from what's actually there.
You give it your URL. It works through your pages — landing pages, blog posts, documentation — and builds an understanding of what you do, which problems you address, and how you describe them. Then it produces seed keywords grouped by theme and by search intent, so informational terms stay separate from commercial ones.
The prompt includes one constraint that matters: don't invent keywords we have no content for. Without it you get an aspirational list — terms for products you don't sell and problems you don't solve. Those terms feel exciting and they're useless, because you have nothing to rank with. The value of this task is that the output is grounded in evidence.
For each cluster it also flags who's likely competing, which is where this hands off to the analysis phase.
Use it to find your real competitors, then run the standard playbook.
Search a handful of terms from each cluster and record what comes back. The domains appearing repeatedly across your clusters are your SEO competitors — that's the empirical definition, and it's often a surprising list. Some names you expected won't be there. Some you'd never considered will be everywhere.
Now the tools become useful. Put those domains into Ahrefs or Semrush and do the gap analysis properly, on the right set of sites. The sequence is what changes, not the toolkit.
If you want to go further upstream and understand the market itself rather than just the search landscape, our guide on using AI for market research and competitive analysis covers that side of it.
No. It replaces the blank document you'd otherwise start from.
Keyword tools are irreplaceable for volume data, difficulty scores, and SERP history. None of that can be inferred from reading your content, and you shouldn't try. But those tools all require you to type something into the search box first, and their suggestion engines expand outward from whatever you give them. Start with a weak seed and you get a large volume of weak suggestions.
This task produces the input. The tools do what they're good at afterward.
The seed list matters more, not less.
Traditional SEO rewards ranking for a term. Generative engines like ChatGPT and Perplexity synthesize answers from sources, so what matters is whether your content is topically credible enough to be cited on a subject. That shifts the unit of value from individual keywords toward coherent topical coverage — which is exactly what a themed seed list gives you and what a flat list of high-volume terms does not.
We wrote more about how that changes optimization in our guide to GEO.
Then this task has less to read and the list will be thin. Run it anyway on what exists, since a small grounded list beats a large invented one, and re-run it once you've published more.
No. It reads your content, it doesn't query keyword databases. Take the seed list into a keyword tool for volume and difficulty.
ChatGPT will generate keywords for a company like yours based on its training data. Sai opens your actual site and reads your actual pages, so the output reflects what you've published rather than what's typical for your category.
Yes. Point it at their URL and you'll get the same treatment — the themes their content covers, grouped by intent. That's a useful second run once you know who your competitors really are.
No. It generates the seed list from your content. Ranking data comes from your keyword tool or Search Console in the next step.
