Keyword Research Scraper – Google, YouTube, Amazon & App Store

This keyword research scraper pulls autocomplete keyword suggestions from seven engines in one run: Google, YouTube, Bing, Amazon, the App Store, DuckDuckGo and eBay. Type one seed such as ai tools and get back hundreds of long-tail keyword ideas that real people are typing right now, as flat rows you can drop into a spreadsheet, a content plan or an LLM pipeline.

Three things set this keyword suggestions scraper apart from every other keyword research scraper on the Store:

• No API key, no login, no proxy. Every source is a public autocomplete endpoint, so a run with the default input finishes in seconds and costs almost nothing in compute.
• Built-in expansion. Alphabet soup (seed a … seed z), question words (how seed, why seed …), prepositions (seed for, seed vs …) and an optional depth-2 pass that feeds every discovered keyword back in as a new query.
• Pay only for keywords you receive. One flat price per delivered keyword row. Sources that return nothing cost nothing.

Why autocomplete is the best free keyword source

Autocomplete is not a guess. Google, YouTube, Amazon and the App Store only suggest completions that enough people have actually searched for, ranked by how often they search for them. That makes a Google autocomplete scraper the fastest way to find:

• long-tail keywords with clear intent (ai tools for small business owners),
• questions people ask (how ai tools work, what ai tools are free for students) – the free alternative to scraping People Also Ask,
• comparison and commercial intent (notion vs obsidian, ai tools like chatgpt),
• platform-specific demand – what people search on YouTube is not what they search on Amazon or in the App Store, and this actor shows all of them side by side.

The alphabet soup method is how every keyword research tool from Ubersuggest to KeywordTool.io gets its data: send the seed plus each letter of the alphabet and collect what comes back. Twenty-six extra queries per seed typically turn 10 suggestions into 200–500 unique keywords. Question and preposition expansion do the same for informational and comparison intent.

What you get

One row per keyword, identical fields whichever source it came from:

Field	Description
keyword	The suggested keyword or phrase
source	google, youtube, bing, amazon, appstore, duckduckgo or ebay
seed	The seed keyword this suggestion came from
modifier	The expansion token that produced it (how, for, a …) or null for the bare seed
query	The exact query string sent to the engine, e.g. how ai tools
level	1 = from the seed or its modifiers, 2 = from the depth-2 re-expansion
position	Rank inside the engine's answer for that query (1 = top suggestion)
relevance	Google's own relevance score for the suggestion (Google only, null elsewhere)
language, country	The locale the query was made in
url	The exact request URL, so every row is reproducible
scrapedAt	ISO 8601 UTC timestamp

Keywords are de-duplicated per source, and the seed itself is never returned as a result. A source and seed that yields nothing produces a single row with an error field explaining why, so the run never quietly succeeds with an empty dataset.

Search volume is not included. Autocomplete endpoints do not expose it, and any actor that claims to get volume from autocomplete is estimating. Use position and Google's relevance as a demand signal, and pair this actor with a volume tool (Google Keyword Planner, DataForSEO or a volume actor on the Store) for the shortlist you actually want numbers on.

Input

The default input is enough to try the keyword research scraper – it researches ai tools on six sources with question expansion and returns roughly 500 keywords in a few seconds.

{
  "seedKeywords": ["ai tools"],
  "sources": ["google", "youtube", "bing", "amazon", "appstore", "duckduckgo"],
  "language": "en",
  "country": "us",
  "expandQuestions": true,
  "expandPrepositions": false,
  "expandAlphabet": false,
  "expandDigits": false,
  "depth": 1,
  "maxKeywordsPerSeed": 500
}

A full content-planning run for a YouTube creator:

{
  "seedKeywords": ["ai for finance", "fp&a", "financial modeling"],
  "sources": ["google", "youtube"],
  "expandQuestions": true,
  "expandPrepositions": true,
  "expandAlphabet": true,
  "depth": 2,
  "maxKeywordsPerSeed": 2000
}

An Amazon seller checking a niche in three marketplaces means three runs (one country each), or one run per marketplace on a schedule:

{ "seedKeywords": ["yoga mat"], "sources": ["amazon", "ebay"], "country": "de", "expandAlphabet": true }

Options

• Seed keywords – one per line. Each seed is sent to every selected source.
• Sources – any mix of google, youtube, bing, amazon, appstore, duckduckgo, ebay.
• Language / Country – language drives Google, YouTube, Bing and DuckDuckGo (en, de, sv, ja …). country picks the Google/Bing/DuckDuckGo region and also selects the Amazon marketplace (22 supported, from us and gb to jp, in, br, ae), the App Store storefront (50+) and the eBay site (19). An unsupported country for those three falls back to the US with a warning in the log – the run still completes.
• Expand with question words – prepends who / what / why / how / when / where / which / can / does / is / are / should / will. 13 extra queries per seed and source.
• Expand with prepositions – appends for / with / without / near / to / like / vs / versus / and / or. 10 extra queries.
• Expand with a-z – appends every letter. 26 extra queries. This is where the bulk of long-tail keyword ideas come from.
• Expand with 0-9 – appends every digit. Good for products, years and versions.
• Depth – 2 sends every level-1 keyword back to the same source as a new query until the cap is hit. Multiplies results roughly 5–10x.
• Maximum keywords per seed – per seed and source. This is your cost ceiling.

Request count per source and seed = 1 + (13 if questions) + (10 if prepositions) + (26 if alphabet) + (10 if digits), plus up to the cap at depth 2. The actor runs eight requests in parallel with small pauses, retries politely on 429/5xx, and isolates every source so one engine being down never affects the others.

Use cases

• SEO agencies and content marketers – run the keyword research scraper on a client's core terms, build topic clusters from Google and Bing suggestions, then map questions to FAQ sections and H2s. Export to Google Sheets and hand the writer a brief.
• YouTube creators – YouTube autocomplete is the closest thing to a free YouTube keyword tool. Research seed + questions on the youtube source, sort by position, and you have a month of video titles people are already searching for. This actor was built by a finance YouTuber for exactly that workflow.
• Amazon and eBay sellers – Amazon keyword research for listings and PPC. Amazon's completion API returns the terms shoppers type in that marketplace, in that language, ranked by demand.
• App developers (ASO) – App Store keyword research: the appstore source returns what people type into the App Store search box, including competitor app names, for 50+ storefronts.
• PPC teams – generate negative keyword lists fast. Alphabet expansion surfaces the free, crack, jobs and salary variants you want excluded from a commercial campaign.
• AI content pipelines and RAG builders – flat rows with source, modifier and level are ready for an LLM to cluster into intents, or for an agent to call through MCP.
• Market researchers – use one keyword research scraper to compare how demand for the same seed differs across Google, YouTube and Amazon, or across country values, without paying for a seat on an enterprise SEO suite.

Pricing

Pay per event: the keyword research scraper charges per keyword row delivered to the dataset. Sources or seeds that return nothing cost nothing, error rows are never charged, and you can cap spend per seed and source with maxKeywordsPerSeed. There is no actor-start fee and no separate charge per source – a keyword from Amazon costs the same as one from Google.

A typical run of one seed on six sources with question expansion returns ~500 keywords. The same seed with alphabet and prepositions on Google and YouTube at depth 2 usually returns 2,000–4,000.

Integrations

Every row from the keyword research scraper is flat JSON, so the dataset drops straight into Google Sheets, Airtable, a CSV export, n8n, Make, Zapier or an LLM prompt. Run it on a schedule and diff the results to see new keywords appear week over week.

Python

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("nourishing_courier/keyword-research-scraper").call(run_input={
    "seedKeywords": ["ai tools"],
    "sources": ["google", "youtube"],
    "expandAlphabet": True,
})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(row["source"], row["position"], row["keyword"])

JavaScript

import { ApifyClient } from "apify-client";

const client = new ApifyClient({ token: "YOUR_APIFY_TOKEN" });
const run = await client.actor("nourishing_courier/keyword-research-scraper").call({
  seedKeywords: ["yoga mat"],
  sources: ["amazon", "ebay"],
  country: "gb",
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.map((r) => r.keyword));

curl

curl -X POST "https://api.apify.com/v2/acts/nourishing_courier~keyword-research-scraper/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN&format=csv" \
  -H "Content-Type: application/json" \
  -d '{"seedKeywords": ["fp&a"], "sources": ["google"], "expandQuestions": true}'

n8n / Make / Zapier – use the Apify node or module, pick this actor, paste the JSON input above and connect the dataset to a Google Sheets "append rows" step. MCP – expose the actor through Apify's MCP server and an AI agent can run keyword research as a tool call.

Limits and fair use

This keyword research scraper reads the same public autocomplete endpoints a browser calls when you type into a search box; no personal data is collected, only the suggestion strings. Each engine returns roughly 8–15 suggestions per query (Bing up to 25), so the expansion options are how you get volume. Very large depth-2 runs across many seeds send thousands of requests; the actor spaces them out and retries on rate limits, and the Proxy option in Advanced is there if you ever need it, but none of the sources need one at normal volumes. Autocomplete answers change daily and by locale – that is the point – so scrapedAt, language and country are on every row. Search volume, CPC and difficulty are not part of what these endpoints expose and are not claimed.

FAQ

Is this keyword research scraper free?

You pay per keyword delivered, nothing else, which makes it a near-free Google autocomplete scraper for small jobs. Apify's free plan credit covers thousands of keywords per month, and there is no API key to buy from Google, Amazon or anyone else.

How many keywords do I get per seed?

Bare seed with question expansion on six sources: about 500. Add alphabet, prepositions and depth 2 on Google and YouTube and expect 2,000–4,000 per seed. maxKeywordsPerSeed caps it.

Does it return search volume?

No. Autocomplete does not expose volume, so this actor does not invent it. It returns each engine's own ranking (position) and Google's relevance score, which is the demand signal Google itself uses to order suggestions. Feed the shortlist to a volume tool if you need numbers.

Can I scrape YouTube keyword suggestions without an API key?

Yes. The youtube source uses YouTube's public suggest endpoint, no Data API quota involved. It works with any language and country.

Which Amazon marketplaces are supported?

US, UK, DE, FR, IT, ES, CA, JP, IN, AU, SE, NL, MX, BR, SG, AE, SA, PL, TR, EG and BE, chosen by the country field. Each marketplace answers in its own language.

Does it work for App Store ASO keyword research?

Yes. The appstore source returns App Store search hints for 50+ storefronts, including the competitor app names people type. It is the same endpoint the App Store app uses.

What is the alphabet soup method?

Append every letter a–z to your seed and collect the autocomplete for each. ai tools a gives ai tools and automation, ai tools b gives ai tools for business and so on. Turn on Expand with a-z and the actor does it for every seed and source.

How is this different from People Also Ask scrapers?

PAA scrapers need a full search results page per query, which is slow and usually needs a proxy. Question expansion here queries the autocomplete endpoint directly with how, why, what … in front of the seed, so you get real question keywords in seconds with no proxy.

Does it work with n8n, Make and Google Sheets?

Yes. The output is flat rows, so any Apify integration can push it to a sheet, a database or a Slack message with no transformation step.

Will one source being down break the keyword research scraper run?

No. Every source and seed pair is isolated. If Amazon times out, the Google, YouTube and Bing rows are still delivered, and the Amazon pair gets one explanatory row instead of silently vanishing.