3.5.2 Search Trend Analysis

In Brief

Description

Search trend analysis reads demand signal out of what people search for — both in traditional search engines and in AI assistants. This is one of the Data Mining methods: you work existing query data rather than running new interviews or surveys. Billions of people use search engines and LLMs every day to express their needs, problems, and interests. Each query is a small signal of demand. Aggregated, these signals reveal patterns that can inform early-stage product decisions, and aggregated search data has been shown to track real-world economic activity ahead of official statistics.

Unlike surveys or interviews, search data captures what people actually do rather than what they say they do. A person searching “how to fix leaky faucet” has a real, immediate problem. A rising trend in “AI writing assistant” indicates growing market interest. A seasonal spike in “tax software” reveals timing opportunities. What “search” means is also widening: even outside Google, a large share of search activity happens on platforms like YouTube, Amazon, Reddit, and Wikipedia. Treat search as a behavior that shows up wherever your audience hunts for answers, not just on one engine.

The method treats two surfaces as parallel data sources:

• Traditional search-engine trends. Trend and keyword-volume tools show how often people type queries into search engines, where they are typed, and how that volume changes over time. Regional engines (Baidu in China, Yandex in Russia) matter where the audience is concentrated outside Google’s footprint.
• AI question trends. A growing share of problem-aware queries are now addressed to AI assistants instead of search engines. People phrase questions to an assistant differently than they search Google — fuller sentences, more context, more comparative phrasing (“which is better for X”). AI-question-trend tooling is emerging: a small set of platforms track which queries surface in AI-mediated answers, which sources the assistants cite, and how visibility shifts over time. Public AI-question datasets are still thin — a lot of the practical work is direct experimentation, asking the same question across multiple assistants and noting how each answers and what each cites.

The goal is not to validate a specific hypothesis but to map patterns of demand and discover opportunities you may not have considered. Comparing the two surfaces is itself diagnostic: if a topic has high search volume but no AI-question presence, the problem is mature and AI assistants may be eating into the channel; if a topic has heavy AI-question activity but flat search volume, you may be early to a shift in how the audience finds answers.

How to

Prep AI Prompt

1. Define your topic area. Pick 3–5 seed keywords related to the problem space. Use both broad terms (“project management”) and specific terms (“kanban board for remote teams”). Include terms your customers might use, not just industry jargon.

2. Expand seed terms across both surfaces. Search-engine queries are short and keyword-shaped (“project management remote team”); LLM queries are longer and conversational (“what’s the best way to run sprint planning when half my team is async”). Generate variants of each seed for each surface — they are not the same list. AI is useful here: you can ask an LLM to rewrite each seed as five plausible Google queries and five plausible AI-assistant-style questions a real founder, ops lead, or end customer would ask.

3. Set up your toolkit. For traditional: open Google Trends and a keyword-volume tool. For AI-question signal: pick at least one AI-search monitoring tool and at least two LLMs you can prompt directly.

Analysis AI Prompt

1. Cluster the keyword landscape. Organize findings into clusters by buyer awareness stage: problem-aware (“how to reduce churn”), solution-aware (“customer retention software”), and brand-aware comparison (“Intercom vs Zendesk”). Do this for both surfaces — the cluster shape often differs between Google and LLMs.

2. Compare the two surfaces side by side. For each cluster, note: traditional search volume and trend direction, AI-question prevalence (cited frequency, AI Overview presence), and which sources the LLMs default to. Topics where both are rising are the strongest demand candidates. Topics where AI-question signal leads but search is flat may be a shift to watch. Topics where search leads but AI assistants don’t engage may indicate a transactional or branded surface that LLMs aren’t yet eating into.

3. Synthesize and generate hypotheses. Combine findings into a summary that answers: What demand exists across both surfaces? Is it growing? What language do customers use in each channel? What adjacent opportunities exist? Use these hypotheses as input for further research such as Customer Discovery Interviews.

When reading the results, keep these interpretation guidelines in mind:

• A sustained upward trend over 2+ years on either surface suggests growing demand rather than a short-lived fad.
• A flat or declining traditional search trend does not necessarily mean no opportunity — the market may be mature, the audience may have moved to AI assistants, or customers may use different terminology.
• Heavy AI-question presence with thin search volume often indicates an early shift in how the audience finds answers. The audience may be there; the channel has changed.
• High search volume with few quality results on the search results page suggests an underserved market on the traditional side.
• Rising “how to” and “best” queries — and conversational “which should I” questions to LLMs — indicate a market where customers are actively seeking solutions.
• Geographic concentration may indicate cultural, regulatory, or infrastructure factors worth investigating.
• Seasonal patterns can inform launch timing and marketing strategy.