Google Trends — Ask What the World Is Actually Searching For

August 18, 2026

Every other search tool in Alfrada OS tells you what has been published. Google Trends tells you what people are looking for — and whether that's more or less than last quarter, in which countries, and what they type alongside it. Ask "is ozempic still trending?" or "compare chatgpt, claude and gemini over five years" and you get a real interest curve, a per-term readout, and a CSV you can hand straight to Code Executor or Report Generator.

What you can do

• Interest over time — google_trends returns Google's 0–100 relative interest index for a term, week by week (or hour by hour on short ranges). Alfrada OS adds a per-term summary on top: average, latest value, the peak and the date it happened, and a rising / flat / falling read of the second half against the first.
• Compare up to 5 terms on one scale. Put them in a single comma-separated query — "chatgpt, claude, gemini" — and Google normalises all of them against the same 100. Each term gets its own colour, its own line, and its own summary row.
• Interest by region — data_type: "GEO_MAP" ranks countries by default, or drops to states/provinces (region: "REGION"), US metros (DMA), or cities (CITY) when you scope to one country. Zero-interest regions are dropped rather than padding the list.
• Related and rising searches — RELATED_QUERIES and RELATED_TOPICS return both the established top list and the rising list, where values are percentage growth rather than the index, and Breakout means growth over 5000%.
• Nine time ranges plus custom windows — now 1-H, now 4-H, now 1-d, now 7-d, today 1-m, today 3-m, today 12-m (the default), today 5-y, all (back to 2004), or an explicit YYYY-MM-DD YYYY-MM-DD window.
• Five Google surfaces — measure web search (default), Images, News, Shopping, or YouTube search interest with property. YouTube interest for a music act and web interest for the same act are genuinely different curves.
• Every result saves as a CSV. Time series and region breakdowns write trends_<type>_<terms>_<timestamp>.csv into the session out/ folder and index it, so Code Executor can regress it, Excel Ops can join it, and Report Generator can cite it — without a second API call. Pass save_data: false to skip.
• A chart in the chat, not a wall of numbers. The result card draws the interest curve with a hover crosshair and per-date tooltip, labels each line at its end, and lists the summary numbers underneath. Region results render as ranked bars; related searches render as top / rising columns that link out to Google Trends.

Where this shows up

• Sizing a market before you build the deck. "Compare search interest in 'electric SUV' and 'plug-in hybrid' over the past 5 years, then chart it." One call, one CSV, and the presentation tools pick it up in the same turn.
• Checking whether the hype is real. "Is 'agentic AI' actually growing, or is it just my feed?" The rising/falling read and the peak date answer that in one line.
• Choosing which market to enter first. "Which German states search most for ski holidays?" Region breakdown at state level, ranked.
• Finding the long tail before a campaign. "What are the rising searches around protein powder in the last 3 months?" Breakout terms are where the demand moved before anyone wrote about it.
• Timing a launch. "When in the year does 'tax software' peak?" Five years of seasonality in one curve.

Try it

• "is ozempic still trending?"
• "compare search interest in chatgpt, claude and gemini over the past 5 years"
• "which countries search most for 'remote work visa'?"
• "what are the rising searches around protein powder in the last 3 months"
• "show me YouTube search interest for lo-fi beats over the past 12 months"
• "seasonality of 'tax software' searches, past 5 years, US only"
• "which German states search most for ski holidays"

Sizing a market with it

Trends emits an index, not units, so it can't size a market alone — but it gets there by anchoring. You supply one number whose absolute value you already know (your own analytics figure, a public company's reported segment revenue, or a cited published stat); Trends measures the ratio between that anchor and your target in a single comparison call, so both sit on one scale; then you convert with a conversion rate and a real price pulled from live listings. Split the total by geography using the regional shares from one region call, and where a public comparable exists, regress its reported quarterly revenue on its own Trends curve to fit the search-to-revenue relationship rather than assume it.

Alfrada OS ships this as a playbook, so you don't have to specify the method: ask for a market size and the agent runs the chain, states the anchor, the ratio and every multiplier, and returns a low/base/high range with the assumptions in a spreadsheet you can edit. It will also tell you when the method doesn't apply — no credible anchor, a term too small to register, or a category that's bought without being searched.

The agent also knows what to pair it with: Google News for why a curve moved, Company Financials for calibration, Google Shopping / Amazon for price and supply, LinkedIn Jobs for whether rivals are committing to a market, Code Executor and Excel Ops for the model, and Task Scheduler to turn a one-off read into a demand history nobody else has.

Heads up

• It's an index, not a volume. Every number is 0–100 relative to the peak of that exact comparison — those terms, that range, that region, that surface. A term scoring 8 next to chatgpt may score 100 on its own. Never compare numbers across two separate calls, and never read the index as searches-per-month. Alfrada OS states this in the tool result and on the card so the agent doesn't launder it into a volume claim.
• Worldwide is the default, deliberately. Unlike Google News and Google Search, Trends does not auto-localise to your detected country — a demand question is usually a global question. Name a place ("in Germany", "US only") and it scopes; say nothing and you get worldwide.
• Five terms is Google's ceiling. Extra terms are dropped and the result says so.
• Low-volume terms come back empty. That's Google filtering, not a failure. Broaden the term, lengthen the range, or go worldwide.
• Same rate as the rest of the SearchAPI family. One Trends call costs the same as one Google News or Google Search call.
• Paste-a-URL works. Drop a trends.google.com link into chat and the tool auto-activates.