Long Tail Keyword Generator — Google, YouTube & Amazon Autocomplete Research

Give it one seed phrase, get back hundreds of real long-tail keywords and questions — pulled live from the same autocomplete indexes Google, YouTube and Amazon use in their own search boxes.

A live run on protein powder returned 1,084 unique keywords in 2.1 seconds: 315 of them questions, 836 of them four words or longer.

Nothing here is invented or LLM-generated. Every keyword is something a search engine volunteered because people actually type it.

How it works

Autocomplete only answers exactly what you ask, so this asks many slightly different things:

• Alphabet soup — appends a through z, so Google reveals its top completion for each letter
• Questions — prefixes how, what, why, when, where, who, can, does, is…
• Comparisons — appends vs, versus, alternative, better than…
• Prepositions — for, with, without, near, under…
• Commercial / intent — best, cheap, buy, price, review, near me…

Results are de-duplicated across every query and source. Keywords that surfaced from several queries are ranked first, because that repetition is itself a relevance signal.

Why not just pay for a keyword tool

AnswerThePublic and KeywordTool.io sell this for $89–$99 per month. This is the same mechanism, priced per keyword, with no subscription and an API you can call from a script or a schedule.

Input

{
  "keywords": ["protein powder", "home espresso"],
  "sources": ["google", "youtube", "amazon"],
  "modifierGroups": ["bare", "questions", "comparisons", "prepositions", "alphabet"],
  "language": "en",
  "country": "us",
  "questionsOnly": false,
  "minWordCount": 4
}

Sources: Google for general intent, YouTube for video topics, Amazon for buying intent. Using more than one also tells you which keywords appear on several platforms.

Localisation: language and country matter a lot — autocomplete is heavily localised. ko/kr, de/de, es/mx all return genuinely different keyword sets.

Output

{
  "keyword": "protein powder vs creatine",
  "seed": "protein powder",
  "sources": ["google", "youtube"],
  "modifierGroups": ["comparisons", "alphabet"],
  "seenCount": 6,
  "relevance": 601,
  "wordCount": 4,
  "characterCount": 26,
  "isQuestion": false,
  "isLongTail": true
}

• seenCount — how many separate queries surfaced this keyword. The best single ranking signal here.
• relevance — Google's own suggestion relevance score, when it volunteers one.
• isQuestion / isLongTail — ready-made filters for content planning.

Common uses

• Content planning — questionsOnly: true gives you a blog or FAQ outline in one run.
• SEO keyword discovery — find the long-tail terms with real demand that competitors have not written for.
• YouTube titles and tags — use the youtube source; its index reflects video search, not web search.
• Amazon product research — the amazon source is pure buying intent.
• PPC negative keywords — everything irrelevant that still autocompletes is a negative to add.
• Localised research — run the same seed across countries and compare.

Pricing

Pay per keyword delivered. Keywords removed by your filters, and failed queries, cost nothing.

Starting a run costs $0.00004 — the platform's $0.00001 minimum, charged once per GB of memory, and these Actors run on 4 GB. That is four cents per thousand runs, and it is the only charge not tied to a row you received.

FAQ

Does this give search volume? No. It gives real keywords plus corroboration and Google's relevance ordering. Volume needs a paid data provider; this is the discovery half, and it is the half that gets you the long tail.

Are the keywords real? Yes — they come from live autocomplete indexes, not a language model.

Does it need an API key? No.

Can I run it on a schedule? Yes, via Apify Schedules, webhooks, or the API. Also available over MCP for AI agents.