Two keywords, both position 3. One sits under a clean list of links; the other sits under an AI overview, a featured snippet and a People Also Ask box, somewhere below the fold. Same rank, different planet.
Tracking positions without tracking the furniture misses half the story of every SERP. Here is the feature layer: what each piece does to clicks, how to track presence, and the two features you can realistically win.
The furniture inventory
| Feature | What it does to clicks | Can you win it? |
|---|---|---|
| Featured snippet | Absorbs a big share, or is position 0 if it is yours | Yes, from page one |
| People Also Ask | Splits attention into question rabbit holes | Yes, per question |
| AI overview | Answers simple queries outright; cites sources | Partly: be a source |
| Local / map pack | Owns local-intent clicks above organics | Via business profile, not pages |
| Image / video rows | Diverts visual intents | With media assets |
| Sitelinks | Amplifies the winner's footprint | Indirectly, via structure |
Track presence, not just position
The tracking unit is the pair: your position plus the furniture above it. Three sources build it.
The quarterly SERP audit: search your money keywords in a private window and log the features present, ten seconds each. Boring, and it is the ground truth everything else approximates.
Search Console's search-appearance filter shows traffic from some feature types, and feature-aware rank trackers log per-keyword furniture automatically, which turns "the SERP changed" from a hunch into a dated event.
That date matters because of the signature it explains: position stable, clicks falling. Half the time that signature is a feature arriving above you, and knowing when it landed separates a snippet problem from a furniture problem.
Winning the snippet
Snippets are drawn from page-one results, which makes every ranking of 2 to 5 under someone else's snippet a takeover candidate.
The block that wins reads almost mechanical: the question as a heading, the direct answer in the first sentences, the nuance after. Your page keeps its depth; the extractable block is the doorway.
Winning the questions
People Also Ask boxes are won one question at a time: a question-phrased heading with a concise answer beneath it, which the question-mining and FAQ structure work you already do produces as a byproduct.
PAA wins compound oddly: appearing for one question keeps you in the box's rotation as it unfolds, so a page answering a cluster of related questions can own several slots of the same feature.
Living with AI overviews
Where an AI overview fully answers the query, some clicks are gone, and the honest response has two lanes. Be quotable: overviews cite sources, and tight, well-structured answers get cited, which preserves presence and some clicks. And re-weight toward surviving intents: the comparison, the walkthrough, the decision-stage query where nobody stops at a summary, the same re-weighting logic as any SERP-reshaping response.
Track it like weather, not like an insult: which of your keywords carry overviews, whether you are cited, and how CTR moved when the overview arrived. The winner-reading habit applies here too: cited pages share traits, and they are copyable.
The one-line takeaway: track the furniture with the position: audit money-keyword SERPs quarterly, date the feature arrivals, take over snippets by out-extracting the current holder, win PAA slots with question-shaped blocks, and answer AI overviews by being quotable where summaries win and deeper where they cannot.