SERP features

A SERP — search engine results page — used to be a predictable layout: 10 organic blue links, maybe some ads at the top. That layout still exists, but it’s no longer the typical case. Most commercial queries now return a layered SERP with multiple feature blocks competing for real estate above and around the organic results.

SERP features are those non-organic blocks. Some are algorithmic (AI Overviews, featured snippets, People Also Ask), some are entity-driven (Knowledge Panel, Local Pack), some are media-driven (Image Pack, Video Pack), and some are commerce-driven (Shopping results, reviews carousel). Each has its own ranking signals, its own appearance rules, and its own optimization playbook.

Three features dominate the visibility conversation for local businesses and AEO-focused operators:

• Local Pack — the 3-business listing block plus map that appears for local-intent queries. The single highest-value SERP block for service-area businesses; we cover its mechanics, ranking factors, and the difference from the expanded finder view in the dedicated entry.
• Knowledge Panel — the entity-driven card Google returns for branded queries. Owning a Knowledge Panel is the clearest signal that Google treats your business as a distinct entity and is willing to surface canonical information about you directly.
• AI Overviews — Google’s AI-generated summary block at the top of an increasing share of SERPs. Often displaces above-fold organic results entirely, so optimizing for the cited source set inside the Overview is its own discipline (see also AEO).

Featured Snippets, People Also Ask, Image Pack, and the reviews carousel matter too — they’re category-dependent and we surface them in the per-query SERP feature detail when they appear.

Every tracked query hits the SERP API, the response is parsed for feature blocks, and the per-feature appearance is recorded for that query × date. Our SERP-fetch path is cached in the dfseo:serp:google namespace per P-12 so multiple orgs querying the same keyword + location on the same day reuse a single upstream call — fast, cheap, and consistent across the dashboards different customers see.

The AI Overviews surface gets special handling: the Google AIO provider in backend/app/core/ai_rank/providers/google_aio_provider.py parses ai_overview.references from the serializer response and exposes each cited URL as a first-class data point. That means your dashboard reports both “did this query trigger an Overview?” and “was my domain in the Overview’s cited source set?” as separable signals.

Different ranking signals, different optimization tactics. Organic rank is a single 1-100 position per keyword based on Google’s core algorithm weighting backlinks, content, technical signals, and user behavior. SERP feature appearance is feature-specific: Local Pack ranks on proximity + GBP signals; Knowledge Panel ownership requires entity-resolution success; AI Overview citation requires being a credible, extractable source on the underlying query.

Concretely, a brand can rank #1 organically and still be invisible above the fold because the AI Overview, Local Pack, and reviews carousel together pushed organic results below the fold. Tracking just organic rank misses that the user never sees the result. Tracking SERP feature appearance separately is what tells you when the above-fold layout has changed against you.

Two trajectories are reshaping the SERP today:

• AI Mode expansion. Google is progressively shifting more queries — especially long-tail and conversational — into an AI-first layout that resembles the Overview but takes more of the page. As that share grows, the AEO-side measurement model (mention + citation rank) becomes the dominant visibility signal for an increasing share of queries.
• Search Generative Experience (SGE) successors. The original SGE label has been folded into AI Overviews and AI Mode, but the underlying direction is the same: more synthesized answer, less link list. Tracking should treat the AI-surface area as a single rising-tide feature set rather than chasing each branded rollout separately.

The practical implication: the SERP feature inventory isn’t stable. A tracker that bakes in a fixed feature list will miss the next 18 months of layout changes. We treat feature detection as a streaming problem rather than a fixed schema — add the parser the moment a new block ships, backfill the trend going forward.