TL;DR: There is no documented Google Autocomplete API. The endpoint everyone copies from Stack Overflow, suggestqueries.google.com/complete/search, is one Google called "unpublished" and "unsupported" when it restricted outside use in 2015. It still answers. I ran one seed keyword across Google, YouTube, Bing and Google Shopping: 299 suggestions in 6 seconds, and the phrases that show up on every engine are the ones worth writing about. Code and output below.
"Where is the documentation?"
That is the literal title of a Stack Overflow question from 24 February 2011. It had 93,109 views on 22 November 2026. The answer is still the same: there isn't any.
Google said it plainly on 24 July 2015. The post calls it a non-official, non-published API and announced it "will be restricting unauthorized access to the unpublished autocomplete API" from 10 August 2015. Google's reason: completions are meant to sit next to search results, and outside that they add little.
So what do you actually get today?
| Question | Answer (checked 22 Nov 2026) |
|---|---|
| Official, documented Google autocomplete API? | No |
Does suggestqueries.google.com still return suggestions? | Yes, from Apify servers, for Google, YouTube (ds=yt) and Shopping (ds=sh) |
| Rate limits, SLA, terms for this use? | Not published. Google can change or block it any time |
| Is one call enough for keyword research? | No. One call gives about 10 suggestions. You need A–Z, question and preposition expansions |
One keyword, four engines
I ran "standing desk" with the seed plus question expansions (how, what, why, is, are...), US English:
| Engine | Suggestions |
|---|---|
| 151 | |
| Bing | 80 |
| YouTube | 52 |
| Google Shopping | 16 |
| Total | 299 (243 unique) |
Run time 6 seconds, cost $0.075.
The interesting part is the overlap. These came up on all four engines:
- standing desk chair
- standing desk amazon
- standing desk with drawers
And these on three: legs, frame, topper, converter, with storage, "does standing desk help", "is standing desk healthy", "are standing desks worth it", "are standing desks good for you".
A phrase that people type into Google, Bing and YouTube is real demand, not one engine's quirk. That list is a content brief.
The top questions, in Google's order: "does standing desk help", "is standing desk healthy", "are standing desks worth it", "are standing desks good for you", "what standing desk to buy". Hello, FAQ section.
The code
I didn't call Google from my own machine. Hammering an unsupported endpoint from one IP gets you rate-limited fast. This runs on Apify instead (apify-client 3.x):
import os from apify_client import ApifyClient client = ApifyClient(os.environ["APIFY_TOKEN"]) run = client.actor("kuezi/google-autocomplete-scraper").call(run_input={ "keywords": ["standing desk"], "engines": ["google", "youtube", "bing", "google_shopping"], "expansions": ["seed", "questions"], "language": "en", "country": "us", }) rows = list(client.dataset(run.default_dataset_id).iterate_items()) print(len(rows), "suggestions") for r in rows[:10]: print(r["engine"], r["position"], r["suggestion"]) summary = client.key_value_store(run.default_key_value_store_id).get_record("SUMMARY") print(summary["value"]["keywords"][0]["crossEngine"][:5]) Real output (first lines):
299 suggestions google 1 standing desk reviews google 2 standing desk converter google 3 standing desk with drawers google 4 standing desk topper google 5 standing desk with storage Each row has the engine, the expanded query, the suggestion, its position, and for Google a relevance score and suggestion type. The SUMMARY record holds the cross-engine list and the top questions.
Google Autocomplete Scraper on Apify. Disclosure: I built this. It charges $0.00025 per suggestion ($0.25 per 1,000), a bit less on paid Apify plans, and it also covers Yahoo, DuckDuckGo and Wikipedia.
Limits, honestly
- It's unofficial. Google never documented the endpoint and can change it. If it breaks, every tool built on it breaks, mine included.
- Suggestions are not search volume. They tell you what people type, not how many. Pair them with Google Trends for direction, or a paid keyword tool for volume.
- Results move. Suggestions depend on country, language and date. Re-run before you publish a brief.
- Shopping is thin. 16 suggestions for a product keyword. Use it for buyer phrases, not coverage.
If you know of an official replacement I missed, tell me in the comments and I'll update this.
AI disclosure: I wrote this with AI assistance. Every quote, number and output above was checked or run on 22 November 2026.

