The Problem Nobody Talks About After the Rankings Report Looks Fine
You’re ranking on page one. Traffic is coming in. Then you actually look at which queries are driving that traffic in Search Console, and the top performers aren’t what you optimized for. They’re close — same semantic neighborhood — but not the same thing. Different intent, different conversion rate, different everything.
That’s Google’s synonym expansion layer doing exactly what it’s designed to do. The problem is it’s doing it to your page, not just to someone else’s.
What Synonym Expansion Actually Is (and Isn’t)
Google has been doing synonym matching since at least the 2013 Hummingbird update, but the mechanism has changed substantially since then. It’s not a lookup table of word substitutions. It’s a vector-space inference system: when a user types a query, the system maps it into a semantic space and matches it against pages whose content lands in the same region, even if the surface-level words are completely different.
This is why a page about “car insurance rates” can rank for “how much does auto coverage cost” without ever using those words. The model has learned, from billions of queries and clicks, that these concepts co-occur in the same decision-making context.
Useful, right up until two semantically adjacent concepts get collapsed into the same vector cluster even though they represent genuinely different tasks for the user.
An Example Worth Spelling Out
Say you write a detailed guide on “SaaS churn rate benchmarks.” Your intent is to serve someone trying to contextualize their own churn — a VP of Product asking “is 5% monthly churn normal for our stage?”
Google’s synonym layer may also surface that page for queries like “how to reduce churn” or “churn rate formula.” Adjacent, yes. Same intent? No. The first is diagnostic. The second is remediation. The third is definitional. Three different jobs. One page can’t do all three with equal depth, and Google will eventually figure that out — usually by watching pogo-sticking patterns back to the SERP for the queries that weren’t really matched.
You can’t see pogo-sticking directly in GSC. But you can see its downstream effect: impressions climbing while average position stagnates, or CTR dropping for specific query clusters in the performance report.
Why This Is Harder to Diagnose Than Cannibalization
Keyword cannibalization — two URLs splitting SERP equity for the same keyword — is at least visible. Two URLs, same keyword, you can spot it in a ranking report.
Synonym drift is subtler. One URL, showing up in the wrong semantic cluster. Your page didn’t change. Google’s interpretation of it did, or the query volume distribution shifted and now a different type of searcher is finding you.
The audit gap is that most SEO tools report which queries a page ranks for but not how well that page’s content actually serves those queries. That second evaluation requires you to segment GSC performance data by query intent — which means doing the intent classification yourself, because no tool automates this reliably.
How to Actually Audit This
Pull 90 days of query data from Search Console for any page you’re investigating. Export it. Cluster those queries by task, not by topic:
- Informational/definitional — user wants an explanation
- Comparative — user is evaluating options
- Procedural — user wants step-by-step guidance
- Diagnostic — user has a specific situation and wants analysis
- Transactional — user is ready to act
For each cluster, calculate average CTR separately. A well-matched page should show roughly uniform CTR across intent types, adjusting for SERP feature interference. If one cluster’s CTR is running at half the average, that cluster is finding your page through synonym expansion and the content isn’t satisfying it.
Most audits skip this entirely. They look at total impressions and total clicks, not at CTR variance across intent clusters. That’s where the signal lives.
Three Ways to Respond (They’re Not Interchangeable)
1. Add a Dedicated Section That Explicitly Serves the Drifted Query
If the volume on the synonym-drifted queries is meaningful, the simplest response is to earn the ranking rather than fight it. Add a section to the existing page that genuinely addresses the adjacent intent. Use a clear <h2> or <h3> that mirrors the drifted query’s language, and give it real depth. Passage indexing means Google can promote a subsection of your page for a specific query without ranking the whole page — you can exploit that deliberately.
2. Create a Separate Page With Signal-Appropriate Anchor Text
If the intents are genuinely incompatible — the definitional searcher and the remediation searcher want completely different things — you want separation. Create a second page that owns the adjacent query, then internally link between them with anchor text that clearly differentiates the two destinations. “SaaS churn benchmarks” linking to “how to reduce SaaS churn” with that exact anchor text pushes Google toward treating them as distinct documents covering distinct jobs.
Don’t use generic anchors like “learn more” or “read our guide here.” The anchor is part of the ranking signal for the destination page. Leave it vague and you’re handing the disambiguation back to Google, which means it’ll probably keep collapsing them anyway.
3. Accept the Drift and Optimize the On-Page Experience for It
Sometimes the synonym-drifted traffic converts better than your original target. You write something for one audience, and a slightly different audience finds it more useful. If CTR and engagement on the drifted queries are strong, don’t fight it — make sure the page’s title tag, meta description, and above-the-fold content don’t create a mismatch between what those users expect and what they land on. A title optimized for “benchmarks” but content that opens on process guidance will create a pogo-stick problem regardless of the page’s overall quality.
Your Keyword Map Probably Has This Problem Everywhere
If you built a keyword map and matched pages to keywords, you almost certainly have synonym drift happening across multiple pages. The map-to-page matching process assumes Google will interpret your target keyword the way you interpret it. Often it will. Sometimes it won’t — especially for terms where user intent has shifted, where AI Overviews have changed which queries surface actual URLs, or where your content has enough topical breadth that Google routes multiple adjacent intents through it.
The fix isn’t rebuilding the map. It’s adding a layer: after you’ve published and started collecting GSC data, come back to every important page and run the intent-cluster CTR analysis above. Quarterly, not one-time.
Start with your highest-impression pages that aren’t your highest-traffic pages. That gap is usually where synonym drift is hiding.