Keyword, Topic & Query Research
Someone types "hp gaming murah" into Google, then opens ChatGPT and asks which one lasts longest for a student budget. Same buyer, same afternoon, two completely different phrasings — and most keyword lists only contain the first one.
The Challenge Brands Face
Search behaviour split in two, and keyword tooling stayed on one side of the split. Google queries are compressed: two to four words, no question mark, often no verb. Assistant queries are the opposite — long, conversational, full of context the person would never type into a search box ("for a student budget", "yang aman buat kulit sensitif"). Both express the same demand. Only one of them shows up in a volume export.
The second gap is harder to see. Search volume tells you a question gets asked, never whether it gets answered well. Plenty of high-volume queries already have five solid articles competing for them, while the questions your buyers actually stall on — the comparison nobody has written, the objection everyone dodges — carry low volume precisely because no good answer exists to click on. Volume-ranked lists systematically bury exactly the topics worth writing.
Assistants have become a research surface, not a novelty, and the phrasing people use there is observable today. A query set built only from search-volume exports now describes a shrinking share of how buyers actually look for things — and the answer gaps it hides are the cheapest visibility available, because no competitor has written into them yet.
Our Approach
Collect the phrasing people really use
We gather queries in the form they are actually written — short informal Indonesian for Google, full conversational questions for ChatGPT, Perplexity, Gemini, and Claude — rather than the polished phrasing marketers invent. That distinction matters: a query set calibrated to real typing behaviour produces a realistic map, while one written in marketing language produces a map of nobody.
Cluster into topics and tag the intent
Related queries are grouped into topic clusters, and every query carries an intent tag and a funnel stage — discovery, consideration, or decision. Clustering is what lets you plan a pillar page and its supporting articles as one structure instead of writing one post at a time, and the funnel tags stop a content calendar from filling entirely with top-of-funnel explainers.
Flag the gaps and rank the queue
We check what AI engines currently answer for each question and mark the weak ones, then sort everything into Trending Now, Evergreen Answers, and AI Answer Gaps. Each topic arrives with its exact query, intent, and a why-now signal, so a writer can pick up the top item and start without a briefing call.
What You Receive
How We Prove Progress
During the optimization period — before conversions fully materialize — these are the metrics we use as proof that the strategy is moving in the right direction.
Who This Is Designed For
This service works best for:
- Content teams that run out of topics by the third week of every quarter
- Brands whose keyword research predates ChatGPT being a research tool
- Founders who want to write less and have each piece aimed at real demand
- UMKM competing in crowded categories where the obvious keywords are already taken
What to Expect
Your writers get a queue of topics backed by real demand on both Google and AI assistants, with the weakest-answered questions at the top.
A first query set lands within about a week once we know your category and competitors, and refreshes on whatever cycle you publish on.
Questions About Keyword, Topic & Query Research
What is query research and how is it different from keyword research?
Query research covers the full natural-language questions people ask AI assistants, while keyword research covers the short phrases they type into a search box. "Skincare lokal aman untuk kulit sensitif" is a query; "skincare lokal" is a keyword. Both come from the same buyer, and a content plan built on only one of them addresses half the demand. We build a single set covering both.
How do you find what people ask ChatGPT about my category?
We build a category query set the way a buyer would phrase it, run it through the assistants, and read what comes back. Doing that across ChatGPT, Perplexity, Gemini, and Claude shows which questions produce confident answers, which produce vague ones, and which name your competitors instead of you. The vague and competitor-heavy ones are the openings.
How many topics do I get per cycle?
Around ten researched topics per cycle, each with its exact query, search intent, funnel stage, and a why-now signal, grouped into Trending Now, Evergreen Answers, and AI Answer Gaps. Ten is deliberate: it is enough to keep a content team busy for a cycle and small enough that every item has been checked rather than bulk-exported.
Does this replace my existing keyword tool?
It replaces the decision, not necessarily the tool. Volume data is still useful as one input, and you can keep exporting it. What changes is what ranks the queue: instead of sorting by volume, topics are sorted by intent, funnel stage, and whether a good answer already exists — which regularly pushes a low-volume question with no decent answer above a high-volume one with five.
Interested? Let's Talk.
We start with a short conversation — no template pitch, no generic proposal. We want to understand your brand first.
Or WhatsApp us: +62 819-7712-1092