Keyword Suggestions — Google, YouTube & Amazon Autocomplete

Give it seed keywords and get every phrase that Google, YouTube, Google Shopping, Amazon, Bing and DuckDuckGo autocomplete suggests for them — asked again with a–z, 0–9 or question words added to reach the long tail — as one clean, deduplicated row per keyword, with the sources that suggested it, its position and its word count. $0.50 per 1,000 unique keywords, no login, no API key.

Run it again — or put it on a weekly schedule — and every row also tells you whether the keyword is new since the last run (isNewSinceLastRun) and how its position moved (positionChange). Autocomplete only shows today's list; this is how you catch a phrase that has just started to appear.

What you get

Field	What it is
seed, keyword	your seed, and the suggested phrase. Duplicates across sources and queries are merged (ignoring case and extra spaces); the first spelling seen is kept. The seed itself is never returned
sources	every service that suggested the phrase, e.g. ["google","youtube","amazon"]. Phrases several services agree on are easy to spot
bestPosition, foundVia	the best position (1 = first suggestion) the phrase had in a suggestion list, and the query that produced it — the seed itself, or something like habit tracker t. The typed text that services echo back as their first line is not counted as a suggestion
wordCount, isQuestion, containsSeed	number of words; true when it starts with a question word or ends with ?; false for containsSeed when autocomplete drifted off topic
isNewSinceLastRun, previousBestPosition, positionChange, previousRunAt	the change since your previous run for the same seed (null on the first run). positionChange is positive when the phrase moved up
country, language, observedAt	the market asked and when

Rows come out ranked: phrases suggested by the most sources first, then by position. Rows with type: "error" carry seed, source, message and failedRequests when a source could not be read; they are free.

Use it for

• SEO and content planning — the long tail around a topic, with the questions people type, from three or more search boxes at once
• YouTube and Amazon research — what viewers search on YouTube and what shoppers type into the Amazon search box of each marketplace, side by side
• Ads and listings — phrases for titles, tags and bid lists, flagged by how many sources agree
• Trend watching — schedule it weekly and filter on isNewSinceLastRun to see phrases that just appeared

Input

{
  "keywords": ["habit tracker", "budget app"],
  "sources": ["google", "youtube", "amazon"],
  "expand": "alphabet",
  "country": "us",
  "language": "en",
  "maxKeywordsPerSeed": 500
}

• sources — google, youtube, google-shopping, amazon, bing, duckduckgo. Default: Google, YouTube, Amazon. Amazon follows country and covers us, gb, de, fr, it, es, ca, jp, au, mx, br, nl, se, pl, tr, ae, sa and sg.
• expand — none (just the seed, about 10 phrases per source), alphabet (the seed plus a to z and 0 to 9: 37 queries per source — the default), or questions (how, what, why, where, when, which, who, can, best, is in front and vs, for, without, near me behind: 15 queries per source).
• country, language — the market, e.g. us/en, de/de, jp/ja, br/pt-BR. Google, YouTube, Bing and DuckDuckGo are asked for that market, and Amazon uses that country's marketplace.
• maxKeywordsPerSeed — default 500. If there are more, the ones suggested by the most sources and ranked highest are kept; only returned rows are charged.

What changed since the last run is remembered per seed in a named key-value store in your Apify account, so the second run of the same input already has the flags. Schedule it (Apify → Schedules) for a regular history.

Pricing

$0.50 per 1,000 unique keywords ($0.0005 each), plus $0.01 per run. Error rows are free. The $0.01 run-start fee is the one charge every run pays, even one that returns nothing; put all your seeds in one run and you pay it once.

One seed with the defaults (Google + YouTube + Amazon, alphabet) returns up to 500 keywords in roughly 10 to 15 seconds of fetching, so about $0.26 including the run fee. Use "expand": "none" for only the top phrases.

Notes

• Public data only: the suggestions anyone sees while typing into those search boxes, without signing in. They show what people type, not how often — there is no search volume.
• Positions are positions inside one suggestion list (the first real suggestion is 1; the typed text that every service lists first is not a suggestion and takes no slot). Adding a letter pushes a phrase towards the top of that query's list, so compare bestPosition between runs (positionChange) rather than between unrelated rows.
• isNewSinceLastRun, previousBestPosition, positionChange and previousRunAt are null on the first run. They are also null the first time you change the seed's country, language, expand or sources, because that is a different list. previousBestPosition and positionChange are null for a phrase that is new. Suggestions fluctuate from day to day, so a few phrases are new on almost every run.
• Your history holds every phrase seen, also the ones cut by maxKeywordsPerSeed, so a phrase that only moved up into the returned list is not reported as new. If a source is down during a run, the phrases it gave you earlier are kept too, so they are not reported as new when it comes back.
• Each source is asked at most about 5 times per second. A request that is rate limited or fails to connect is retried twice (after 5 and 10 seconds). A source that still fails gets a free error row for that seed, and one that fails three requests in a row is left out for the rest of that run. Results from the requests that worked are always included. If nothing at all could be read, the run fails instead of finishing empty.
• Amazon has one suggestion service per marketplace. India is not offered: its service did not answer when checked. Asking for Amazon in a country without one skips Amazon with a free error row — or refuses the run if Amazon was the only source. Google Play has no public suggestion service, so it is not a source.
• Question words in questions mode are English; for other languages use alphabet. isQuestion recognises English, German, French, Spanish, Italian, Portuguese and Dutch question words, and a trailing question mark in any language. It is a simple rule, so a phrase like "do it yourself desk" counts as a question.
• wordCount counts space-separated words. Japanese, Chinese and Thai are written without spaces, so there it counts the space-separated parts. For seeds in those scripts containsSeed ignores spaces.
• No personal contact data: a suggestion that is an email address or a phone number is dropped.
• Suggestions in other languages come back as written, accents and scripts intact (checked on Google in 14 languages, and on Amazon, Bing and DuckDuckGo in Japanese and German).

To track where videos rank for these keywords, see YouTube Search Rank Tracker. To see which apps rank for them, see App Store Scraper.

For AI agents

{"keywords": ["habit tracker"], "sources": ["google"], "expand": "none"} returns about 9 rows, one per suggested keyword. Rows have type: keyword or error. For the long tail use "expand": "alphabet"; run the same input again later and filter on isNewSinceLastRun: true for phrases that appeared in between.

**Something broke or looks wrong?** Open an issue on this Actor's Issues tab and I'll look at it. **Found it useful?** A short review on the Store page helps other people find it.