You've picked a keyword, you're ready to brief a piece of content, and there's an easy mistake waiting right at the start, writing the wrong kind of page for a job nobody searching that phrase actually wants done. Knowing how to do a SERP analysis before you write a single word is what keeps that from happening. It's the step where you actually look at what Google is already serving for your target keyword and figure out what the searcher wants, what format wins, and what a new page would have to do to earn its place there.
This guide walks through a SERP analysis guide you can run on one keyword in under an hour, the kind you'd use right before handing a brief to a writer or a production workflow. It's scoped narrowly on purpose. You're not mining a hundred keywords for content ideas here, and you're not doing a company-wide gap analysis against a competitor's whole site. You're reading one page-one result set for one keyword, and deciding what to build.
What a SERP analysis actually tells you
A SERP is just the page of results Google shows for a search, short for search engine results page. When you're analyzing search results for content strategy, you're not counting how many blue links show up. You're reading the page for evidence: who's winning, in what format, and why. That evidence tells you three things before you write anything. What the searcher is actually trying to do. What kind of page and presentation the results reward. And what would have to be true of your page for it to earn a spot next to, or above, what's already there.
It's worth being clear about what this step doesn't promise. Reading the SERP well doesn't guarantee your page will rank, but it gives you a defensible, evidence-based starting point instead of a guess, and that's a real difference in how a brief turns out.
Step 1: Fix the query and viewing conditions
Before you look at a single result, write down exactly what you're searching, and under what conditions. The exact target keyword, who it's for, which country or region matters, the language, the device, and the date you're doing this. Then actually search that phrase and look at the real page. Don't lean on a keyword tool's intent label or a screenshot someone sent you last quarter.
If you're signed into Google, there's a "Try without" option that shows results without your personal search history factored in. There's also a personalized recommendations setting you can check. Note whether you used either one. And if most of your readers are on mobile, look at the mobile results. The desktop layout can look meaningfully different.
Done when: someone else could run the same search, in the same market, on the same device, and see roughly what you saw.
Where people go wrong: treating an incognito window as some kind of neutral, universal view. Google is upfront that results can still shift based on your location, your device, and what else is trending, personalization setting or not. A SERP is a snapshot of one moment, not a fixed ranking you can bank on next month.
Step 2: Inventory the page before you click anything
Scan the page top to bottom before you open a single result. Note the ads separately from the organic listings. Write down anything else that shows up: an AI Overview, a featured snippet, a People Also Ask box, videos, images, a local map pack, roughly where each one sits on the page. Then sort out which of those items actually answer the query directly, and which point somewhere else entirely.
For the organic results that are actually relevant, capture the title, the source domain, and what kind of page it looks like before you even click through. A screenshot is a nice backup, but your written inventory should stand on its own without one.
Done when: you can point to which results are the real organic competition, separate from the features and paid placements sitting above or around them.
Where people go wrong: assuming a People Also Ask box means your article needs a giant FAQ section, or treating an AI Overview as proof you'll get cited in it. Google has said plainly that AI Overviews show up only when its systems judge them genuinely additive, and often they don't show up at all. A featured snippet works the same way. Google picks it. You can write a clear, useful answer, but you can't mark a page "make this the snippet."
Common mistake: pulling every visible SERP feature into your outline as if each one is a required section. A People Also Ask question only belongs in your piece if answering it actually helps the reader finish the task your page is for.
Step 3: State the dominant intent, and name any split
Now put the query wording together with the pattern you're actually seeing in the results, and ask one question: what does the searcher need to do after they read or use this result? Learn something. Compare their options. Complete a transaction. Reach a specific site. Those four buckets, informational, commercial, transactional, and navigational, are the standard shorthand, and write your answer as one plain sentence. Something like, "the searcher wants a sequence they can follow to judge this keyword before they brief an article."
If you're seeing two genuinely different clusters of results, that's a real secondary intent, not noise. Write it down along with the evidence for it. And don't assume a query starting with "how" can only pull in conventional articles. Plenty of "how" queries return tools, calculators, or videos instead.
Done when: you and your team could both say, out loud, what job this reader is trying to get done, and explain any competing read on the query using what's actually on the page.
Where people go wrong: taking an SEO platform's single intent label at face value and never checking it against the live SERP. Mixed intent is common. Those four labels are a starting vocabulary, not a diagnosis that settles the question for you.
Step 4: Separate the winning type, format, and angle
For each relevant organic result, write down three separate things: its content type (article, video, product page, tool), its format (how-to, listicle, comparison, review), and its angle, meaning the specific promise or focus it's making, like "beginner-friendly" or "tested firsthand." These aren't the same thing, and folding them together loses the decision you actually need to make. Look for the pattern that repeats, and note any meaningful exception to it.
Then ask the harder question: does the article you're planning actually fit that pattern well enough to do the reader's job? If most of the relevant results are tools, tests, or product pages rather than text, a written article might just be the wrong asset for this exact keyword, no matter how well you write it.
Done when: you can finish the sentence "make this type of page, in this format, with this angle," and say plainly where the SERP is mixed enough that you're not fully sure.
Where people go wrong: treating "type" and "format" as interchangeable, or slapping the current year onto a title because competitors did. A date on a page should mean something changed. Changing the number in the headline without changing the content underneath isn't a content quality strategy, and readers notice.
Pro tip: when nine of ten results for a query turn out to be interactive tools instead of articles, as one well-known SEO case study found for a query about testing English level, that's the SERP telling you the expected asset isn't a blog post at all. Treat a pattern like that as a signal specific to its own keyword, not a number to expect on yours.
Step 5: Open the pages, and separate requirements from repetition
This is the step people skip, and it's the one that matters most. Read the strongest, most directly relevant organic pages in full, not just their title tags. For each one, note the opening answer, the major sections, the actual method or deliverable it gives the reader, how much detail it goes into, whether it shows firsthand experience, how it sources its claims, and any visuals or tools that make the task easier. Also note where a reader might still be stuck after finishing the page.
Then sort what you found into two piles. What's genuinely necessary coverage, the stuff every useful page on this topic has to include, and what's just repetition, headings that competitors happen to share without it meaning much. Only pull a People Also Ask question into your own outline if answering it actually moves your specific page toward finishing the reader's task.
Done when: you can name the coverage every page in the pile needs to include, and you can point to at least one specific, honest place where the existing pages fall short. If you genuinely can't find one, say that instead of inventing a gap.
Where people go wrong: stacking every competitor's headings into one bloated outline, repeating claims you haven't checked, or calling a random missing tangent a "gap" just because nobody covers it. Google's own guidance for people-first content asks whether a page brings original analysis and offers real value beyond what's already ranking. It also asks whether the page shows genuine expertise and actually leaves the reader able to finish what they came to do.
Step 6: Test your proposed improvement against competitive reality
Now turn the same scrutiny on yourself. Are the pages you just read genuinely useful and directly relevant, or are their examples vague, their instructions incomplete, their claims out of date? And separately: can your team actually supply something stronger? A firsthand walkthrough, a verified example, a clearer decision rule, a one-page brief a reader could actually use?
If you have access to an SEO platform, page-level referring domains and site-level authority scores are worth a glance here, but only as directional context alongside content quality, never as the deciding factor. A referring domain counts unique linking websites, which isn't the same number as total backlinks. Some pages with thin backlink profiles still rank well because the content itself does the job better. Don't assume the reverse either, that a gap in authority means the keyword is easy.
Done when: you can write the decision in one sentence. "Proceed, because we can add X and Y that this reader actually needs." Or "change the asset type." Or "hold off, because we don't have the expertise or evidence yet to genuinely improve on what's there."
Where people go wrong: treating a proprietary difficulty score or authority number as if it were a Google ranking factor, or as a precise traffic forecast. A domain rating measures the strength of a backlink profile on its own scale. It isn't a ranking factor, and a traffic estimate for the top result isn't a promise of visits to a page you haven't written yet. There's no universal minimum word count, link count, or heading count that earns a spot on page one. Google has said directly that it has no preferred word count.
Step 7: Write the one-page content brief
Everything up to here has been observation. This step turns it into a decision. Pull it into one page: the exact keyword and the search conditions you recorded, the primary reader job and any secondary intent worth naming, the page type, format, and angle you're proposing, the answer your page has to give early, the essential tasks or questions it has to cover, what the competing pages already do well, the specific improvement you're making, and what evidence, examples, or assets someone needs to gather before drafting starts.
Keep observed SERP facts and your own editorial choices clearly separate on the page. A useful test for whether the brief is actually finished: could a writer produce the right piece from it without reopening every competitor page themselves?
This is usually where a marketing lead's job shifts from analysis to production. Once you've made the call on intent, format, and the specific improvement to make, you can carry that decided idea into New Ideas or Planned Content inside Content Studio and let DeepSmith turn the brief into a drafted, on-brand article. DeepSmith doesn't replace the judgment call you just made about this specific SERP. It picks up once you've made it, and takes the idea the rest of the way to a finished draft.
Done when: the brief makes the actual content decision, not just a list of links or an exported SERP report. The improvement you're proposing should be specific enough that someone else could tell whether the finished page delivered on it.
Where people go wrong: turning an observed feature into a guarantee ("there's a featured snippet, so we'll get it too"), drafting something that's really just a rewrite of the top result, or asking a writer to invent expertise, numbers, or outcomes the brief never gave them.
Step 8: Check the finished article against the brief
Before you publish, hold the produced page next to the brief you wrote. Does it deliver on the stated intent? Does it use the format you decided on? Does it give the answer early, cover what you flagged as essential, and actually make the improvement you promised? Ask someone who's never seen the brief whether they could complete the task using only the finished page. Check the facts, the product claims, and the voice, rather than assuming a finished draft automatically passed review.
If the SERP shifted meaningfully in the time between briefing and publishing, that's worth a quick second look at your key assumptions before you hit publish. DeepSmith's Content Studio produces a researched, linked, on-brand draft grounded in your product and voice context through Deep IQ, but that review step against your own brief is still yours to run, because it's what tells you whether the finished page actually did what you set out for it to do.
Done when: the page does the job you set out for it, backs up what it claims, and has a clear reason to exist next to what was already ranking.
Where people go wrong: mistaking "SEO optimized" or "publish-ready" for a ranking guarantee, when a well-structured, well-sourced page still has to earn its place the same way every page on that SERP did.
What to do next
Run this SERP analysis guide once on the next keyword sitting in your backlog, before you write a brief for it. You'll notice pretty quickly whether the keyword wants an article at all, or whether you've been about to write the wrong asset. Once the brief is done and you've made the call on format and improvement, that's the point where a tool like DeepSmith can take over the drafting, linking, and metadata work so you can move to the next keyword instead of the next few hours of writing. You can try it with a 7-day free trial.
