Intent Is Obvious — SERP Analysis Should Decode Competitive Signals

SERP analysis is the practice of reading a results page as a record of what Google has already decided about a query — which formats win, which entities recur, which freshness threshold is enforced. Most people treat SERP analysis and competitive signals as a glorified intent check: informational, commercial, transactional, navigational, done. That's the floor. The ceiling is somewhere else entirely.

Grow and Convert's guide on the topic puts it bluntly — most marketers' SERP analysis "does not dig deep enough." That sentence should be taped to every content brief. Because once you've labeled the intent, you've answered the easy question. You haven't started the hard one.

Intent Classification Is the Floor. The Competitive Signal Is the Ceiling.

Eric Carlson's October 02, 2025 update to Siteimprove's "Search Intent Strategy to Rank Higher and Prove SEO ROI" makes the argument plainly: algorithms surface answers that make a user think "Finally, this page gets me," not pages that mechanically check a jargon box. Fine. Everyone agrees. But agreement at that altitude buys you nothing.

The disconnect is mechanical. "Best project management software" is commercial. So is "best CRM for solo founders." So is "best email tool 2026." Three queries, same label, three completely different SERPs. One is dominated by G2 and Capterra. One is dominated by Reddit threads and indie newsletters. One is half AI Overview, half video. The intent label is identical. The competitive reality is not.

This is where SERP analysis stops being about the user and starts being about the page. What Google rewards in this specific niche, today, at this depth, with these entities, at this freshness — those are competitive signals. Read them or guess. There is no third option.

Neil Patel's writing on latent semantic indexing makes the broader point that Google uses aggregated interaction data to assess relevance. Translation: ranking signals run well past intent-matching into behavioral and authority dimensions a keyword tool will never show you. The SERP is the only place those signals are legible.

What "Competitive Signals" Actually Means on a Results Page

Irina Diaconu's May 19, 2026 piece for the AWR SEO Guide defines SERP analysis as examining top-performing websites and SERP features to understand search intent and the requirements for outranking competitors. The second half of that sentence is the one people skip. Requirements. Not vibes.

Competitive signals are the things the SERP is telling you about the cost of entry. Format requirement. Depth requirement. Entity requirement. Freshness requirement. Media requirement. The PAA box isn't a "nice to have for H2s" — it's a list of subtopics Google has decided are inseparable from the main query. Skip three of them and you've published a page that is, by Google's own published anatomy, incomplete.

Terry C Power's LinkedIn teardown on decoding competitors hammers the same nerve: stop copying competitors, start decoding them. He recommends pulling the top 10 results for 20 money keywords as the first move to identify your real SERP competitors — not the ones your sales team complains about, not the ones in your category report. The ones Google has actually placed next to you. Often a different list.

The signals worth reading: who ranks across multiple queries in your cluster (the structural competitor), what content type dominates position one through five (the format mandate), how recently those pages were updated (the freshness floor), which SERP features eat real estate above the fold (the CTR tax), and what entities show up in headings across every winner (the table stakes).

A Research Workflow That Decodes Instead of Describes

Terry C Power frames the work as a 60-minute teardown per cluster. That's the right order of magnitude — long enough to be honest, short enough to be repeatable. Below is how that hour breaks down when the goal is decoding signals, not restating intent.

Incognito pull: Open the query in a clean session, correct locale, no personalization. Pull the top 10 organic results plus every SERP feature above the fold. Screenshot the whole page. This is the artifact. Everything downstream references it. If the AI Overview eats the first scroll, note it — that's a CTR signal that changes the economics of the entire cluster before you write a word.

Competitor set: From those top 10, identify which domains repeat across the cluster's 15–20 sibling queries. Per Power's method, this is where the real SERP competitor list comes from, not from a brand-tracking spreadsheet. Tools like Ahrefs or Semrush surface the gap and link picture; Google Search Console shows share of voice and cannibalization inside your own property; Screaming Frog handles on-page; AlsoAsked maps the PAA entity web; Wayback shows freshness behavior.

Format and depth audit: Open the top five organic pages. Are they listicles, comparison tables, definitive guides, product pages? What is the median word count, not the average — averages lie when one outlier is a 6,000-word monolith. What H2 subtopics show up in at least three of five winners? Those are mandatory. What shows up in one? Optional. What shows up in none but lives in PAA? Opportunity.

Entity and freshness read: Catalogue the named entities — products, people, datasets, standards — that recur across winners. Then check publish and update dates. If four of five top results were updated within 90 days, the query deserves freshness and your evergreen draft is already losing. If the SERP is stable for two years, freshness is not the lever.

Signal synthesis: Write the brief against the decoded signals, not against the keyword. Format mandate, depth floor, entity list, freshness expectation, media requirement, PAA coverage. Anything in the brief that doesn't trace back to a signal on the screenshot is taste, not strategy. Taste is fine. Just label it as such so the editor knows what's load-bearing and what isn't.

Why Intent Typologies Keep Failing the Brief

The four-intent model — informational, navigational, commercial, transactional — is useful the way a compass is useful. It tells you roughly which way to walk. It does not tell you whether the trail is paved, washed out, or under three feet of snow.

There is an academic analog worth borrowing. The 2019 survey "Neural Approaches to Conversational AI" by Gao, Galley and Li (arXiv:1809.08267) groups conversational systems into three categories — question answering agents, task-oriented dialogue agents, and chatbots. Useful taxonomy. Nobody would build an actual production system from the taxonomy alone. You'd need to know the corpus, the latency budget, the failure modes, the evaluation set. The taxonomy is orientation. The build is everything after.

SERP analysis works the same way. "Commercial intent" tells you to walk toward comparison content. It does not tell you whether the winners use third-party review schema, whether the comparison table needs nine columns or four, whether feature matrices outrank prose, whether the SERP rewards a single-product opinion or a multi-product roundup. Those answers live on the page, not in the label.

This is also where fractured intent — the mixed SERP — stops being a problem and starts being a tell. When Google can't decide, the algorithm is showing you the edges of its confidence. A hub page that satisfies multiple micro-intents is one response. A sharper, narrower page that wins a single sub-intent decisively is another. The SERP tells you which bet is cheaper. The intent label tells you nothing.

What an SERP Analysis Output Should Actually Look Like

Most SERP analyses end with a paragraph about intent and a list of H2s lifted from competitors. That's a description. A decode produces a decision artifact. The difference shows up in what an editor can do with the document.

Signal layer	What weak SERP analysis reports	What a decode produces	Source it draws on
Intent	"Commercial intent — comparison content"	"Commercial, but the top 5 are third-party review sites; first-party product pages do not rank above position 8"	Siteimprove, Oct 02, 2025
Competitor set	"Our usual competitors in the category"	"Top 10 across 20 money keywords surfaces three domains we weren't tracking"	Terry C Power teardown
Format & depth	"Long-form blog post, 2,000 words"	"Median 2,400 words across top 5; all five use a comparison table above the fold; none open with a definition"	AWR SEO Guide, May 19, 2026
Freshness	"Evergreen topic"	"4 of 5 winners updated within 90 days; Wayback shows quarterly refresh cadence"	Nightwatch SERP guide
SERP features	"PAA present"	"AI Overview occupies first scroll; PAA contains 6 questions, 4 absent from any current winner — gap"	CONTADU live SERP method
Behavioral signal	"Not assessed"	"Aggregated interaction data implies dwell-time floor; thin pages in this cluster decay fast"	Neil Patel, LSI piece

A reader can act on the right column. The left column gets re-summarized in the next meeting and forgotten.

How This Maps to the Actual Cost of a Content Program

Nightwatch's SERP analysis guide — the one with live SERP snapshots, historical SERP data, and competitor position tracking — is a reasonable tool layer for the monitoring half of the job. AWR's framing splits the work cleanly: on-demand SERP analysis for keyword research and content briefing, and ongoing SERP monitoring for tracking landscape shifts. Both halves cost time. Neither half is automatable to zero.

A program that takes signals seriously will spend more on research per article than a program that doesn't. That cost shows up in the brief — longer, more specific, more defensible. It shows up in the draft — narrower, denser, less generic. It shows up in the refresh schedule — quarterly, not annually, on any cluster where the SERP told you freshness was the price of entry.

The honest tradeoff: programs that skip the decode publish faster and decay faster. Programs that decode publish slower and hold position longer. Either is a valid bet. Pretending the decode is optional while wondering why pages stall is not.

The Question Worth Asking Before Every Brief

Before any article gets written, the brief should answer one question: what did the SERP tell us that we couldn't have guessed from the keyword? If the answer is "nothing," the analysis didn't happen. The intent label happened. Those are different things.

Intent is obvious. It almost always is. The competitive signal — the format mandate, the entity floor, the freshness tax, the SERP-feature CTR drag, the structural competitor you weren't watching — that's the work. Everything else is filing.

Sources

• Search Intent Strategy to Rank Higher and Prove SEO ROI — Siteimprove, Eric Carlson
• SERP Analysis — AWR SEO Guide, Irina Diaconu
• How to decode your competitors in SEO — Terry C Power
• A Complete Guide to Reading and Using SERP Data — Nightwatch
• SERP Intent Analysis: How to Decode What Google Really Wants — CONTADU
• Neural Approaches to Conversational AI — Gao, Galley, Li (arXiv:1809.08267)

FAQ

What is SERP analysis, really?

SERP analysis is reading a results page as a record of what Google has already decided about a query — which formats win, which entities recur, which freshness threshold is enforced. Used as an intent check, it's filing. Used as a decode of competitive signals, it tells you the actual cost of entry before you write a word.

Why isn't intent classification enough?

Because the label is identical across SERPs that aren't. "Best project management software," "best CRM for solo founders," and "best email tool 2026" are all commercial — but one SERP is G2 and Capterra, one is Reddit and indie newsletters, one is half AI Overview. The label tells you nothing about which game you're actually playing.

What counts as a competitive signal on a SERP?

Format mandate, depth floor, entity list, freshness expectation, media requirement, and PAA coverage — plus the structural competitors who repeat across sibling queries. If an AI Overview eats the first scroll, that's a CTR tax that changes the economics of the entire cluster before anyone drafts a brief.

How long should a proper SERP teardown take?

Roughly an hour per cluster, following Terry C Power's framing — long enough to be honest, short enough to be repeatable. Shorter and you're guessing; longer and you're rationalizing. The hour breaks into incognito pull, competitor set, format and depth audit, entity and freshness read, and signal synthesis.

What should the output of SERP analysis look like?

A decision artifact, not a description. An editor should be able to act on it: format mandate, median word count (not average — averages lie when one outlier is a 6,000-word monolith), mandatory H2s that appear in three of five winners, entity list, freshness cadence from Wayback, and PAA gaps. Anything else is taste.

When is the four-intent model actually useful?

It's a compass — it tells you roughly which direction to walk. It won't tell you whether winners use review schema, whether the comparison table needs nine columns or four, or whether the SERP rewards a single-product opinion over a roundup. Use it for orientation, then throw it away and read the page.