Answers

How can I tell if search intent changed for a keyword I already rank for?

Intent lives in the results, not in the keyword#

You cannot read intent from a keyword string. "CRM software" could reasonably mean a definition, a comparison, or a purchase, and which one it means is decided by what searchers clicked on often enough for the engine to settle on a format.

That's why the diagnostic is compositional. Categorise the current top ten by result type and compare against the same list from when you ranked well:

If your page is a guide and eight of the ten results are now tools or product pages, you're ranking on residual authority against a format the query no longer wants.

The threshold worth applying: three or more of the top ten changing result type is a real shift rather than normal churn. One or two swapping is ordinary volatility and reverses often enough that acting on it wastes a rewrite. Confirm across two consecutive months before committing, because a single capture catches temporary experiments Google runs and then abandons.

Getting the historical comparison#

You probably don't have a SERP snapshot from two years ago. Three substitutes work:

Your own Search Console data. If the queries bringing clicks to the page shifted in character, that reflects the same change. Compare the query mix across two 90-day windows a year apart.

Position and CTR divergence. Holding position while CTR falls means the page is shown and not chosen, which is what happens when your format stops matching expectations.

The Wayback Machine on competing pages. If the current top three were guides when they started ranking and are comparison tables now, they adapted to a shift you didn't.

Four symptoms that point to intent rather than competition#

Position held, CTR fell. The strongest single signal. You're visible and unattractive.

Traffic held, conversions fell. The visitors arriving want something different from what the page offers.

Bounce rate rose with no page change. Same content, worse fit.

The page ranks for a different query set than it used to. The engine reclassified what it's for.

Any two together usually justifies a format change rather than a refresh.

Distinguish intent shift from the alternatives#

Three other causes produce similar-looking symptoms, and treating them as intent shifts wastes a rewrite.

A competitor simply did it better in the same format. Check whether the top results are still the same type as yours. If they are, it's competitive, not intent. See Did a competitor take my traffic by publishing a better answer?.

A SERP feature absorbed the click. An AI Overview or featured snippet answering the query directly reduces CTR without any intent change. See Why is my organic traffic dropping while my rankings stay the same?.

Seasonal intent rotation. Some queries genuinely mean different things at different times of year. Compare against the same period last year before acting.

What AI surfaces add to the picture#

Intent is now expressed twice: in what a search results page returns, and in how an assistant chooses to answer.

Our own state of AI search research, covering more than 22,000 answers, found Google AI Mode returns a citation-rich answer in 97.4% of matched B2B queries and Perplexity in 93.2%. The shape of those answers is a direct read on intent, because a model answering with a ranked list of vendors has interpreted the question commercially, while one answering with a definition has interpreted it informationally.

Reading both signals together resolves ambiguous cases faster than the SERP alone. If the results page is mixed but the AI answer consistently returns vendor comparisons, the commercial reading is winning.

Google's documentation on AI features adds a wrinkle: AI Mode may issue multiple related searches across subtopics, so one query can be interpreted several ways at once. Where that happens, splitting beats trying to satisfy every reading in one URL.

Act on the finding#

Format change, not a refresh. If intent moved from informational to commercial, adding depth won't help. The page has to become the format the query expects.

Consider splitting. One URL serving two intents usually loses both. Splitting beats compromising.

Check the query is still worth it. A shift toward transactional can mean it now belongs to product pages you don't have.

Track SERP composition over time in Analyze AI#

The reason nobody catches intent shifts early is that nobody snapshots SERP composition, so the comparison is unavailable when you need it.

The Google Search node runs a DataForSEO SERP request returning the results and features present for a query. Run on a schedule with workflow-memory available, the agent accumulates its own history, which means in six months you have the comparison you wish you had today.

Prompt Responses adds the AI-surface reading, showing how assistants answer the same question and whether the shape of that answer is drifting.

A monthly agent:

Start (schedule, monthly) → Loop / For Each over your priority query set → Google Search node per query (capture result types and features) → Prompt Responses on the matching prompts → workflow-memory (retrieve the same capture from prior runs) → Code (classify each result into the six intent types, compute the composition shift against prior months) → GSC Page-Keyword Breakdown (position and CTR for your pages on those queries) → Prompt LLM (flag queries where composition moved more than a threshold and your CTR fell) → Excel export → Send Email to the content lead.

workflow-memory is the whole mechanism here. A single SERP capture tells you what intent is today. A series tells you when it changed, which is what lets you act while the position is still recoverable rather than after it's gone.

FAQ#

Related answers#

• Did a competitor take my traffic by publishing a better answer?
• Why is my organic traffic dropping while my rankings stay the same?
• When should I rewrite, merge, redirect, or retire an underperforming page?
• Which pages and queries explain most of my organic decline?