Overview
Google Trends Keywords Research is a google trends scraper: a standalone desktop bot that types one keyword into the public Trends explore page, waits for the related queries widget, and captures the CSV export the page generates, written to related_query.csv.
What does Google Trends Keywords Research do?
You give it a keyword - the configuration ships with a required keyword field prefilled with the example default “ai adoption by industry” - and it does the rest without another click. The launcher loads the Trends explore start URL pinned in its project settings, types your keyword into the explore search input, presses Enter, waits for the related queries export control to appear on the results page, triggers it, and saves the file the platform produces. There is no dataset extraction step and no reformatting: the artifact is the platform’s own CSV, byte-for-byte as generated, landing on your disk as related_query.csv. Because everything runs on your desktop, your research interests - which seeds you are expanding, which categories you are probing, which clients you are building keyword backlogs for - never pass through anyone else’s servers. One honesty note up front: unlike platform-neutral listings in this directory, this bot is platform-specific by design. Its selectors are pinned to current Google Trends markup, which is why it carries experimental stability: pinned selectors are the normal maintenance cost of automating one specific page.
Who uses a google trends scraper?
The SEO freelancer asks “what long-tails cluster around this seed right now?” and needs a rising-queries list per client niche, dated and defensible. The content marketer asks “is this topic heating up before I commission ten articles?” and needs the rising percentages as a directional signal, not a guess. The market researcher asks “what vocabulary does the audience actually use for this category?” and needs the related query text itself - the words people search, not the words the brief assumed. All three want an artifact they produced themselves: a CSV with the keyword and the collection date attached, sitting in their own folder, rather than a screen inside someone else’s subscription.
How does an seo keyword scraper run work, step by step?
Every command in the run is linked to its reference page. First, the launcher navigates to the start URL from the project settings - the Trends explore page with an en-US locale. Then fill types the value of ${keyword} into the explore search input, identified by a data-explore-url attribute selector, clearing any previous content first. press sends Enter to the same input, submitting the search. wait_for_selector then blocks until the export control of the related queries widget - the button carrying widget-name="RELATED_QUERIES" - exists on the page, and wait_for_download triggers that control and captures the generated file to related_query.csv. That is the whole run: four commands, one required input, one output file. The keyword variable is required, so a run without it fails before any navigation rather than silently researching the wrong thing.
What does the related queries export look like?
The bot produces no derived dataset of its own, so the columns below are the platform’s export format, shown as an illustration of a rising-queries file:
| Query | Rising |
|---|---|
| ai adoption statistics | +1,250% |
| generative ai in manufacturing | Breakout |
| ai governance checklist | +350% |
Query is the related search term exactly as the platform lists it. This is the column you actually research with: the vocabulary your audience uses around your seed.
Rising (or a relative top index, depending on the widget state) is the platform’s own magnitude signal. Rising values are growth percentages against the prior period; “Breakout” means growth too large to measure. Neither is an absolute search volume, and the bot never invents one.
The file is written verbatim - no renamed columns, no added fields, no parsed numbers. If you need a numeric column for sorting, split the percentage in your spreadsheet so the transformation is visible and yours.
Can it serve as a keyword research automation tool?
Yes, within its stated scope: it automates the tedious loop of navigating, typing, waiting, and exporting for one seed at a time, and it produces a consistently named, dated artifact you can diff across weeks. What it will not do is replace judgment. Rising percentages are relative, Breakout is unmeasurable, and none of it is volume. The working pattern is a pipeline: run the launcher per seed, store each related_query.csv under a filename carrying the keyword and the date, merge the batch in a spreadsheet, and prune. If you need saved histories, alerts on new risers, or grid-style tracking, that is a product category, not this bot - the comparison below sets out the trade.
Is it a complete trends data download workflow?
For the related queries widget, yes - and that is deliberately the only file it produces. It does not export the interest-over-time chart, related topics, or regional interest breakdowns, and it does not loop over regions or languages: the start URL carries a fixed en-US locale. One run, one seed, one widget, one CSV. Each of those boundaries keeps the run predictable; each is also a documented limit rather than a hidden one.
How does it compare to other ways to get trend keywords?
Copying rows by hand works for one seed and collapses at twenty, and it produces inconsistent files nobody trusts a month later. This bot produces the same deterministic file on your own machine for nothing per run, and the workflow configuration is yours to keep. Against subscription research platforms, the difference is structural: those keep your history in their cloud, estimate volumes with proprietary models, and bill monthly. Here the data lands on your disk and the marginal cost of another seed is your own bandwidth. Against metered API services, the same logic holds - per-call pricing versus a launcher you already own. The trade is real: no dashboards, no stored history, no built-in scheduling. You own the artifact and the workflow outright, and you keep your own records.
How do you set it up on Windows, macOS and Linux?
Download the launcher for your platform from this page. Each download is a single raw file - there is no archive and no extraction step on any platform - and the filename carries the configuration id, so do not rename it. On Windows, run the .exe directly; SmartScreen may ask for confirmation on first launch, and “More info” then “Run anyway” proceeds once you have verified the published checksum. On macOS, two launchers are published - pick the one matching your machine’s chip (Apple menu, About This Mac: Apple Silicon or Intel), then make it executable with chmod +x ./BOT_google-trends-keywords-research.sh, clear the Gatekeeper quarantine with xattr -d com.apple.quarantine /path/to/BOT_google-trends-keywords-research.sh, and launch. On Linux, the launcher is the same single shell script: chmod +x ./BOT_google-trends-keywords-research.sh then run it. The first run presents the required keyword field with its example default. For recurring research, schedule the launcher with Task Scheduler, cron, or launchd - the bot has no internal scheduler.
What do I check when a run downloads no CSV?
In order. First: is a consent or cookie interstitial sitting in front of the explore input? The configuration ships no dismissal step, so the fill times out behind it - this is the most common first-run failure, especially from EU addresses. Second: did the platform present a verification challenge or sign-in wall? No solver is configured, so the run stops honestly instead of fighting it. Third: has the page been redesigned? Both pinned selectors - the data-explore-url input and the RELATED_QUERIES export button - must match live markup, or the wait step times out. Fourth: did the keyword produce an empty or absent related queries widget? The export control never appears and the wait expires. Fifth: ordinary network and page-load problems. The contract is that failure is loud - a timeout and a missing file, never a stale or fabricated spreadsheet.
What are the limits of the bot?
One keyword per run; no region or language loop; related queries only, never interest-over-time, topics, or geography; relative signals, never absolute volumes; no scheduling, alerting, or history - it is a snapshotter. It reads only what a logged-out visitor sees and holds no credential fields. It carries no CAPTCHA solver, so a challenge ends the run. And its selectors are pinned: when the platform changes markup, the bot needs a configuration edit, which is exactly why its stability is marked experimental.
What is responsible trends research?
Automated interaction with the platform may conflict with its terms of service; review them with open eyes before your first run and own that decision. Keep request rates at what a patient human could sustain - a few runs per session, not a loop. The data collected is aggregate and contains no personal data, but the output still carries obligations: do not use rising-query lists to manufacture engagement or spam content at scale, and cite the collection date whenever you present the numbers, because trend signals drift weekly. The acceptable-use policy applies in full before your first run.
