Most keyword research workflows stop at search volume and keyword difficulty. Both numbers assume a roughly uniform SERP — ten blue links, maybe a featured snippet, you rank high enough, you get clicks. That assumption has been wrong for a few years now, and in 2026 it’s actively misleading.
The SERP layout for a given query determines how much organic real estate actually exists before a user makes a decision. That layout varies enormously — not just by industry, but by query modifier, device, location, and whether Google has decided to inject an AI Overview. If you haven’t mapped what Google actually renders for your target queries, you’re optimizing for a slot that might not exist in a useful form.
What’s Actually Competing for the Click
For a query like “best weather API for aviation” — one we track directly — the SERP typically shows an AI Overview pulling from G2 and Reddit threads, a People Also Ask block, two or three aggregator pages, and then standard organic results starting around position 4 visually, which might be position 8 in the raw ranking data. The top-of-page real estate is gone before organic even starts.
For “METAR decoder” it’s different. No AI Overview, no PAA box, just a mix of tool pages and reference docs. Organic position 1 is genuinely position 1 on screen. Click-through rate for rank 1 on that query is probably 3–5x higher than rank 1 on the aviation API query, even if volume is comparable.
KD scores don’t capture any of this. A KD 30 query with an AI Overview eating 40% of the page above the fold can drive less traffic to a rank-3 result than a KD 50 query with a clean organic layout. That inversion is real and it repeats across most competitive niches we’ve audited.
How to Actually Map SERP Features at Scale
The low-tech version: pull your target queries, open incognito, manually note what appears above the fold. Slow, but it builds intuition fast. Do this for twenty queries and you’ll see patterns around which intent types trigger which features.
The scalable version requires tooling. STAT and Semrush both track SERP features by keyword over time — not just whether a feature exists, but whether you own it. That distinction matters. If a featured snippet exists for a query you rank for and you don’t own it, that’s an opportunity with a known mechanism to address. If an AI Overview consistently appears and pulls from sources you’re not cited in, that’s a different problem requiring a different fix: off-site authority, structured data, or both.
A practical audit framework we use:
- Classify each query by SERP layout type — clean organic, AI Overview present, featured snippet present, heavy feature load (PAA plus Shopping plus Local Pack stacking). Four buckets, not a twenty-column spreadsheet.
- Estimate real click opportunity, not just rank — rank 2 on a query with an AI Overview and a PAA block delivers materially lower click share than rank 2 on a clean SERP. SparkToro’s CTR research and Ahrefs’ click data are useful reference points here; treat any specific figures as directional rather than precise, because they shift as Google tests new layouts.
- Re-sort by layout-adjusted opportunity — multiply estimated search volume by estimated click share at your current or target rank. Queries where you sit at rank 4 on a clean SERP frequently outperform queries where you rank 1 on a feature-heavy one.
AI Overviews: When They’re a Threat and When They’re Not
AI Overviews appear most consistently on informational queries where Google is confident it can synthesize a complete answer. “How does PageRank work” will have one. “PageRank checker for site audit” probably won’t, and if it does, it’ll be thin and link-heavy. Transactional and navigational queries tend to have lighter AI Overview presence, or none at all.
The threat is concentrated in informational content. If your strategy relies heavily on top-of-funnel informational traffic — which is a legitimate approach, but know what you’re signing up for — audit how many target queries now have AI Overviews and whether your content is being cited inside them. Getting cited in an AI Overview is not the same as ranking in organic results. It’s citation-based, not position-based, and the signals that drive it overlap significantly with E-E-A-T and schema markup rather than keyword targeting.
If you rank in the top three organically on a query but aren’t being cited in the AI Overview for that query, it usually comes down to one of two things: the model is pulling from sources it treats as more authoritative (Reddit threads, Wikipedia, established publications), or your content doesn’t answer the query in a form the model can excerpt cleanly. Both are fixable, but they require different interventions.
People Also Ask Is an Intent Map, Not Just a Ranking Opportunity
Most SEO advice on PAA focuses on how to rank inside the boxes. Fine, but there’s a more useful framing: PAA boxes are Google showing you the full shape of the user’s intent cluster. The sub-questions it surfaces are queries Google has decided are semantically adjacent to the primary query — effectively its own internal topic model made visible.
If your page targets the head query but ignores the PAA questions, you’re missing the sub-intents Google considers part of the same session. A page that addresses the head query and the three most common PAA questions in a coherent structure will outperform one that ignores them — not because you “answered more questions” but because you aligned with Google’s own model of what a complete answer looks like for that query type.
This connects directly to passage-level indexing. Google can rank individual passages, so a well-structured page that covers PAA sub-questions in distinct, labeled sections is more useful to its retrieval mechanism than a long page covering the same ground in undifferentiated prose.
The Caveat Most SERP Audits Skip
SERP features are personalized and localized to a degree aggregate tools can’t fully capture. A Semrush SERP snapshot reflects a specific data center location with no personalization applied. What your actual users see varies. This matters most for queries with local intent — even queries that don’t look local can trigger Local Packs or localized AI Overviews for users in specific regions.
If your traffic concentrates in particular geographies, check SERPs from those locations directly, not from a default US data center view. A query with a clean organic SERP in one region can have a Local Pack eating the top half in another. You’re not competing for the same real estate, so the same ranking strategy won’t produce the same results.
Where to Start
Pull your top 30 target queries by current organic traffic contribution. For each one, record what SERP features appear, whether you appear in any of them, and whether the layout has shifted in the last six months (STAT’s SERP feature history makes that last part fast). Then rebuild your priority list based on layout-adjusted click opportunity rather than raw ranking position.
The ranking position you hold is not the product. The SERP layout you’re competing inside is. Worth knowing which one you’re actually in before deciding how hard to fight for it.