Keyword Ideas: Google, YouTube, Bing & Amazon Autocomplete

Keyword Ideas is a keyword research tool that collects long-tail keyword ideas and question keywords from the autocomplete of Google, YouTube, Bing and Amazon in a single run. Give it one or more seed keywords. It expands each one with A–Z letters, question words, commercial modifiers (best, top, vs, review, price…), prepositions and your own prefixes or suffixes, merges the suggestions across engines, and tags every keyword with where it was found, the seed, the modifier and its search intent.

• 🔎 4 sources in one run. You get one clean table instead of four separate tools, and you can see which keywords appear on Google and Amazon and YouTube.
• ❓ Question keywords (how, what, why, can, is…) for FAQ sections, People-Also-Ask style content and video ideas.
• 🌍 Any country and language, plus Amazon suggestions from 22 Amazon marketplaces, including amazon.in, amazon.co.uk, amazon.de and amazon.co.jp.
• 🪶 No browser and no API key. It runs on 256 MB of memory. In local tests, one seed on all 4 engines returned 1,517 unique keywords from 196 requests in about 10 seconds.
• 💸 Errors are free. If an engine fails, you get a free error record instead of a failed run.
• 📈 Built-in hand-off to Google Trends. The best keywords are saved as a ready-made input for Google Trends Scraper, so you can check which ideas are actually growing.

What keyword data does it return?

Each dataset item is one unique keyword:

Field	Example	Meaning
keyword	how to clean a coffee maker with vinegar	The autocomplete suggestion
engines	["google", "youtube", "bing"]	Where it was suggested (merged across engines)
engineCount	3	Number of engines that suggested it (a strong demand signal)
seed	coffee maker	Your seed keyword
modifier / modifierType	how / question	What was typed to find it: seed, alphabet, number, question, commercial, preposition, custom or depth
query	how coffee maker	The exact text typed into the search box
position / positions	1 / {"google": 1, "bing": 3}	The best position, and the position per engine
googleRelevance	1250	Google's own suggestion relevance score (when available)
intent	question	question, commercial, navigational or informational
intentSignals	["how"]	The words that decided the intent, so you can re-classify with your own rules
isQuestion, wordCount, containsSeed	true, 8, true	Handy filters for long-tail and FAQ keywords
depth, parentKeyword	2, coffee maker cleaner	For "suggestions of suggestions" (depth 2–3)
country, language, amazonMarketplace	IN, en, amazon.in	The locale that was used

The key-value store also has:

• RUN_SUMMARY: keywords per engine and per intent, keywords unique to each engine, requests, failures, stop reason and per-engine HTTP statistics.
• TRENDS_INPUT: the top 5 keywords per seed, formatted as an input for the Google Trends Scraper.

How to find keyword ideas from autocomplete

1. Click Try for free (or Start).
2. Enter one or more seed keywords, one per line, e.g. coffee maker.
3. Choose the sources (Google, YouTube, Bing, Amazon), the country and the language.
4. Pick the expansions: A–Z, questions, commercial (best/top/vs/review/price), prepositions (for/with/near…), numbers, or your own prefixes and suffixes.
5. Optionally set Depth to 2 or 3 to also expand the strongest new keywords.
6. Click Start, then export to CSV, Excel, JSON or Google Sheets, or read the results through the API.

The prefilled example (coffee maker, capped at 300 keywords) finishes in well under a minute.

Why use this keyword tool?

• Cross-engine demand signal. A keyword that Google, YouTube and Amazon all suggest is a stronger bet than one found on Google alone. engineCount and RUN_SUMMARY.uniqueToEngine show you both the overlap and the hidden gems that only one engine suggests.
• YouTube keyword research and Amazon keyword research in the same table as Google SEO keywords. That is useful for creators, Amazon sellers (listing titles, backend search terms, PPC) and content marketers.
• Intent tags out of the box. Filter intent = question for blog FAQs, or intent = commercial for buyer keywords and affiliate content.
• Polite and reliable. It paces requests per engine (with jitter), retries with exponential backoff, honours Retry-After, and detects blocks (429/403/captcha pages). If an engine starts blocking, that engine switches to Apify Proxy automatically. If an engine keeps failing, it is skipped, and the run still returns everything else.
• Spending cap. Set Max keywords in total or Maximum cost per run, and the Actor stops cleanly at the limit.

Popular use cases

• SEO content planning: turn one topic into hundreds of long-tail keywords and question keywords, then group them into articles.
• YouTube video ideas: see what people actually type into the YouTube search box (engines contains youtube).
• Amazon product research and listing optimisation: find the real search terms on amazon.com, amazon.in, amazon.de and 19 more marketplaces.
• PPC keyword lists: use the commercial modifiers (best, cheap, price, review, vs, alternative) to find buyer-intent terms.
• Local and multilingual SEO: any country and language, e.g. Germany/de or India/hi, plus custom prefixes such as kaise or near me.
• Validate with Google Trends: send the best ideas to Google Trends Scraper (see below).

Input examples

One seed on all engines (the default):

{ "keywords": ["coffee maker"] }

India, several seeds, prepositions on, depth 2:

{
  "keywords": ["masala chai", "air fryer"],
  "country": "IN",
  "language": "en",
  "expandPrepositions": true,
  "depth": 2,
  "depthKeywordsLimit": 10
}

YouTube and Google questions only, long-tail filter:

{
  "keywords": ["home workout"],
  "engines": ["google", "youtube"],
  "expandAlphabet": false,
  "expandCommercial": false,
  "minWordCount": 4
}

German Amazon keywords with custom suffixes:

{
  "keywords": ["kaffeemaschine"],
  "engines": ["amazon"],
  "country": "DE",
  "language": "de",
  "customSuffixes": ["mit mahlwerk", "für 1 person"]
}

How many keywords and requests per run?

Each seed is queried once per modifier per engine. With the defaults (A–Z + 14 question words + 8 commercial modifiers, 4 engines), that is 49 queries × 4 engines = 196 requests per seed. In our local tests, that returned:

Test	Requests	Unique keywords	Time
coffee maker, US/en, all 4 engines, defaults	196	1,517 (Google 669, YouTube 490, Bing 524, Amazon 394; 347 on 2+ engines)	~10 s
kaffeemaschine, DE/de, A–Z + custom prefixes	120	906	~5 s
2 seeds, IN/en, prepositions + depth 2	520	800 (capped at 400 per seed)	~20 s

Real numbers vary by keyword and country. See Cost for runs on the Apify platform.

Send keyword ideas to Google Trends

Autocomplete tells you what people search for. Google Trends tells you whether it is growing and where. When Prepare input for Google Trends is on (the default), the run saves a TRENDS_INPUT record like this:

{
  "searchTerms": ["coffee maker review", "coffee maker machine", "coffee maker with grinder", "coffee maker no plastic", "coffee maker cleaner"],
  "geo": "US",
  "timeRange": "today 12-m"
}

Paste it into Google Trends Scraper & API to get interest over time, interest by region or city, and related rising queries. You can also chain the two Actors automatically:

from apify_client import ApifyClient

client = ApifyClient("<YOUR_API_TOKEN>")
run = client.actor("mitp1997/keyword-ideas-autocomplete").call(run_input={"keywords": ["coffee maker"], "maxItems": 500})
trends_input = client.key_value_store(run["defaultKeyValueStoreId"]).get_record("TRENDS_INPUT")["value"]
trends = client.actor("mitp1997/google-trends-scraper-reliable").call(run_input={
    "searchTerms": trends_input["searchTerms"], "geo": trends_input["geo"], "timeRange": trends_input["timeRange"]})

Use it as a keyword API

Python

from apify_client import ApifyClient

client = ApifyClient("<YOUR_API_TOKEN>")
run = client.actor("mitp1997/keyword-ideas-autocomplete").call(run_input={
    "keywords": ["coffee maker", "espresso machine"],
    "engines": ["google", "amazon"],
    "country": "US",
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    if item["status"] == "ok":
        print(item["keyword"], item["engines"], item["intent"])

JavaScript

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_API_TOKEN>' });
const run = await client.actor('mitp1997/keyword-ideas-autocomplete').call({ keywords: ['coffee maker'], country: 'GB' });
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.filter((i) => i.intent === 'question').map((i) => i.keyword));

curl (run and get the items in one call)

curl -X POST "https://api.apify.com/v2/acts/mitp1997~keyword-ideas-autocomplete/run-sync-get-dataset-items?token=<YOUR_API_TOKEN>" \
  -H 'Content-Type: application/json' \
  -d '{"keywords": ["coffee maker"], "maxItems": 200}'

It also works in n8n, Make and Zapier (Apify app: "Run Actor" and "Get dataset items"), and with AI agents through the Apify MCP server.

Cost

Price: $1 per 1,000 keywords (pay per event, keyword-result = $0.001). Apify platform usage is included in that price, error records are free, and you can cap the spend with Max keywords in total or Maximum cost per run.

The Actor is light: no browser, 256 MB of memory, and small JSON responses (about 260 KB of traffic per seed on all 4 engines). For reference, the platform usage we measured on 23 Oct 2026 (256 MB, no proxy; you don't pay this separately):

Run	Keywords	Run time	Platform usage	Usage per 1,000 keywords
Prefill: coffee maker, US, capped at 300	300	10 s	$0.0018	$0.006
2 German seeds, DE/de, all defaults	2,253	43 s	$0.0121	$0.005
2 seeds, IN, prepositions + depth 2	800	56 s	$0.0050	$0.006

Most of that usage is dataset writes, not compute. Error records (status: "error") are never counted as results.

Errors and blocking

• Each engine has its own pacing, retries and block detection. Failed requests never stop the run.
• No proxy is used by default, because these autocomplete endpoints answer plain HTTP requests. If an engine returns 3 blocked responses in a row, that engine only switches to Apify Proxy (datacenter first, then residential if your plan has it). You can turn this off, or force a proxy for all requests, under Advanced.
• If an engine still fails, you get one free error item per seed and engine: {"status": "error", "seed": "...", "engines": ["amazon"], "error": "...", "failedRequests": 15, "totalRequests": 49}.
• If nothing could be collected, the run is marked as failed, so schedules and monitoring notice. You can turn this off with Fail the run if nothing was found.

FAQ

Is this the same as Google Keyword Planner? No. Autocomplete shows what real users type, but it does not give search volume. Use it to discover ideas, then check volume in Keyword Planner, or check trends with the Google Trends Scraper.

How is intent detected? With transparent word rules (English-oriented): question words → question; best/cheap/price/review/vs/buy/near me… → commercial; login/app/official/.com… → navigational; everything else → informational. The matched words are in intentSignals.

Which Amazon marketplaces are supported? US, CA, MX, BR, GB, IE, DE, FR, IT, ES, NL, BE, SE, PL, TR, AE, SA, EG, IN, JP, AU and SG. For other countries, Amazon suggestions come from amazon.com, and the log says so.

Can I get keywords in Hindi, German, Spanish…? Yes. Set Language (hi, de, es…) and Country. Use Custom prefixes/suffixes for language-specific modifiers, because the built-in question and commercial words are English.

Why do Google and Bing sometimes suggest keywords without my seed? They correct spelling and suggest related brands. Turn on Only keywords containing the seed to drop them.

Does it need a proxy or an API key? No. Proxy is optional and is only used automatically if an engine blocks.

Limitations

• Autocomplete has no search volumes or CPC data.
• Suggestions are personalised and change over time. Results come from a neutral, logged-out session in the selected country and language. They are not tied to a specific city.
• Amazon suggestions come from the marketplace that matches the country, in that marketplace's default language.
• Intent tags are heuristic and mostly English.

Support

Found a bug or need another source (e.g. Pinterest, DuckDuckGo, eBay)? Open an issue on the Issues tab.

More Google data Actors

• Google Trends Scraper: interest over time, by region, related queries and trending searches.
• Google Ads Transparency Scraper: every ad a competitor runs on Google Search, Display and YouTube, with ad text, images, videos, dates and regions.
• Google Jobs Scraper: job listings with salaries and direct apply links.