Documentation Python quickstart Blog Free tools hello@quanticdata.ioLog in

Triller Proxies: 0 Words, Domain for Sale

What triller.co returns to a proxied client by fetch mode and exit country, measured on 28 September 2026: 0 words and 114 bytes over plain HTTP, a rendered domain-for-sale notice of 92 words from the United States, 161 from Germany and 92 from Brazil, and 10 archived snapshots from the Wayback Machine collector in 7.3 seconds
What triller.co returns to a proxied client by fetch mode and exit country, measured on 28 September 2026: 0 words and 114 bytes over plain HTTP, a rendered domain-for-sale notice of 92 words from the United States, 161 from Germany and 92 from Brazil, and 10 archived snapshots from the Wayback Machine collector in 7.3 seconds

There is nothing on triller.co for a proxy to read. On 28 September 2026 the profile URL returned 114 bytes and 0 words over plain HTTP from the United States, Germany and Brazil: an HTML document whose only content is a script that redirects to /lander, which is a listing where the domain is for sale. Rendered, the sale notice arrives in the exit country's language, 92 words in English or Portuguese and 161 in German. The app's online services stopped in 2026. The setting on Triller is to spend nothing, and if you need what Triller used to publish, the Wayback Machine collector returns 10 archived snapshots in 7.3 seconds.

We publish this page because the keyword still gets searched and every answer on it is wrong by omission. The results for "triller proxies" are proxy statements filed by Triller Group Inc. with the SEC, a game trailer and a card-game thread; the results for the website are a 2020 tutorial, APK mirrors and the Wikipedia article, which is the only one that says what happened. A creator-research tool with Triller still in its source list is measuring a parked domain. Here is the measurement.

The profile URL returns 114 bytes and a redirect from every country

We audited triller.co/@triller, the address pattern the app used for public profiles, through residential exits in three countries, once without JavaScript and once rendered.

Exit countryPlain HTTPRendered in a browser
United States0 words, 114 bytes, no title, no canonical92 words, title triller.co is for sale — Get a price in 24 hours, canonical on the registrar's sale page
Germany0 words, 114 bytes, identical161 words, title triller.co steht zum Verkauf
Brazil0 words, 114 bytes, identical92 words, title triller.co está à venda

The plain response is the whole story in one line: an HTML document with no body, only a window.onload handler that sends the browser to /lander. It carries no title, no description, no canonical and no Open Graph tags. Any parser that expects a profile gets an empty document with HTTP 200, which is the shape that silently corrupts a dataset: a scheduled job would record 0 followers, 0 posts, no error.

Following the redirect over plain HTTP from Brazil ends on the registrar's aftermarket listing: 266,749 bytes, 145 extractable words, a Portuguese title and a canonical pointing at the listing. The rendered audits from three countries end on the same listing. There is no profile behind any of it.

The exit country changes the language of the sale notice, and nothing else

This is the one geographic finding, and it is worth stating precisely because it is the pattern you will see on any parked domain. The listing page localises by the visitor's country: English from the United States, German from Germany, Portuguese from Brazil. The German version is longer, 161 words against 92, because the German copy is longer, not because Germany sees more. A monitor that counts words per country would report a 75 percent difference between exits for a page that says the same thing in three languages.

The rule we apply on every target applies here in its purest form: assert on what the page is, not on how many words it has. On TikTok a short page is a regional partition; on Bluesky it is a render that did not hydrate; on Triller it is a domain that no longer hosts the service. The number that tells you which is the canonical, and here it points at a registrar.

robots.txt and llms.txt say the domain is for sale

The robots.txt at triller.co is 66 bytes: User-agent: *, Allow: /, an LLM-Policy: /llms.txt line and a sitemap reference. That is a parking template, not a social platform's crawl policy. The llms.txt it points to is 568 bytes and states, in the registrar's words, that the domain is listed for sale on the aftermarket with buy-now, make-an-offer and lease-to-own options and an escrow of five to seven business days. The domain publishes its own status to crawlers and to language models; there is no ambiguity to resolve.

The Wikipedia article, last modified on 22 September 2026, gives the timeline: from July 2025 the app lost its core functions, including signing in and viewing other users' content, and as of August 2026 all online services in the app no longer function and it is no longer available on the App Store. TrillerTV, the live-sports product that shares the name, is a different site on a different domain and is not what a "triller proxies" search is asking about.

Where Triller's pages still exist: the archive, 10 snapshots in 7.3 seconds

If your job is research on the 2020 short-video market, or verifying that a creator really had a Triller presence, the data is not on triller.co and it is not going to come back. It is in the Internet Archive. We ran the Wayback Machine collector on the domain twice on 28 September 2026.

RunRowsSecondsWhat the captures show
triller.co from 1 January 2020107.3Home page captures from 10 January to 3 April 2020, HTTP 200, 7,421 to 7,608 bytes each
triller.co from 1 January 2026101.5HTTP 404 on every capture from 1 January to 23 March 2026; HTTP 200 again from 5 May 2026 at 2,433 bytes, the parking page

Read the second run as a timeline: the domain answered 404 for the first quarter of 2026 and came back in May as a 2,433-byte page, which is the sale listing. Each row carries the capture timestamp, the original URL, the status, the content digest and a playback URL, so the next step is one plain fetch per snapshot through the archive, not through a proxy pool. With match_prefix the same collector lists captures of every path under the domain, which is how you find archived profile pages if the archive ever captured them.

The collector costs $0.0003 per delivered snapshot, so both runs above cost $0.006 together, and nothing is billed for a run that returns no rows. That is the whole Triller data pipeline in 2026: a list of captures for a fraction of a cent, then archived pages.

What it costs, counting a gigabyte as one billion bytes

How you fetch itOn-wire bytesWhat you getPer GBCost at $0.80/GB
Profile URL, plain HTTP114A redirect script, 0 words~8,770,000 fetches$0.00000009
Redirect target, plain HTTP, Brazilian exit266,749145 words of sale notice in Portuguese~3,749 pages$0.00021
Profile URL, rendered, any exitn/a92 to 161 words of sale noticen/awasted
Wayback Machine collector, 10 snapshotsn/a10 rows with timestamps, status, digest, playback URLn/a$0.003 per run

The instruction is to spend less, and here it is absolute: a tool that still schedules Triller fetches is paying for 114-byte redirects and 266-kilobyte sale notices. Remove the source, and if you need the history, pay $0.0003 per snapshot for the archive index instead. If you would rather run one API for whatever replaces Triller in your list, our web scraping API returns any page as Markdown for $0.0002 without rendering, through the same residential exits.

What is legitimate here, and the one line about terms

There are no platform terms to read: triller.co no longer serves the platform's Terms of Service, only a registrar's listing, and the app's own terms cannot be accepted through a service that does not function. What remains legitimate is archival: research on public pages the archive captured, verifying a creator's historical presence for a portfolio or a due-diligence file, and checking whether links to triller.co in bios and press pages still resolve, which they do not. The domain is on the aftermarket, which means anyone can buy it; a brand that still links to it from its own pages should know that.

Do not try to reach the defunct app's endpoints, do not treat the parking page's HTTP 200 as a live source, and do not resell archived personal data: archived posts remain their authors' content and, in Europe, personal data, whatever happened to the platform that hosted them.

The setting that works on Triller

  • Network: none for the site itself. If you fetch it anyway, residential Basic at $0.80/GB returned the redirect on the first request from three countries, at 114 bytes; there is nothing that needs a household address.
  • Fetch mode: engine: tls, plain HTTP, only to confirm the state: 0 words and a redirect script. Do not render; the rendered page is 92 to 161 words of sale notice.
  • Country to pin: none. The exit changes the language of the notice, 92 words in English or Portuguese and 161 in German, not the content.
  • When the proxy is not enough: it is not the tool here. The Wayback Machine collector returned 10 snapshots in 7.3 seconds from 2020 and 10 in 1.5 seconds from 2026, at $0.0003 per snapshot; for platforms that still publish, the web scraping API at $0.0002 per page.
  • Free tier: every account gets $2 of free API usage per month, which covers about 6,600 archived snapshots.

Sources & further reading

FAQ

Quick answers on triller proxies.

Something else? Ask us →

Can I scrape Triller profiles through a proxy?

No, because there are none to scrape. On 28 September 2026 triller.co/@triller returned 114 bytes and 0 words over plain HTTP from the United States, Germany and Brazil: a script redirecting to a page where the domain is for sale. Rendered, the same URL is a 92-word sale notice in English, 161 words in German, 92 in Portuguese.

Is the Triller app shut down?

Its online services are. The Wikipedia article, revised 22 September 2026, states that from July 2025 the app lost signing in and viewing other users' content and that as of August 2026 all online services no longer function and it is not on the App Store. The domain's own llms.txt, 568 bytes, states it is listed for sale on the registrar's aftermarket.

Does the exit country change what triller.co returns?

Only the language of the sale notice. The plain response was 114 identical bytes from three countries; the rendered notice was 92 words in English from the United States, 161 words in German from Germany and 92 words in Portuguese from Brazil, all with the same canonical on the registrar's listing. A word count per country would report a 75 percent difference for the same page.

Where can I get historical Triller data?

From the Internet Archive. Our Wayback Machine collector returned 10 home-page captures from 2020 in 7.3 seconds, each HTTP 200 and 7,421 to 7,608 bytes, and 10 captures from 2026 in 1.5 seconds showing HTTP 404 from 1 January to 23 March and a 2,433-byte parking page from 5 May. Each row has a playback URL; the collector costs $0.0003 per snapshot.

What does triller.co robots.txt allow?

Everything, because it is a parking template: 66 bytes with User-agent: *, Allow: /, an LLM-Policy line pointing to /llms.txt and a sitemap reference. There is no platform crawl policy left to follow, and no platform Terms of Service are served on the domain any more.

Why does a Triller fetch still return HTTP 200?

Because the parked page answers 200 with a redirect script, and the sale listing it leads to answers 200 as well: 266,749 bytes and 145 words from a Brazilian exit. A monitor that checks status alone records a live source; the canonical, which points at the registrar, and the 0-word body are the signals that tell you the platform is gone.

Measure before you fetch, and archive what is gone

The Wayback Machine collector returns a domain's capture history at $0.0003 per snapshot, and every account gets $2 of free API usage per month. For the platforms that still publish, residential exits from $0.80/GB and one key for proxies, scraping and collectors.

Related reading

Social proxiesPlurk Proxies: 118 Words Without a Browser

One public Plurk profile measured through residential exits in Taiwan, Japan and the United States on 28 September 2026. Over plain HTTP it returns 118 words in 32,781 bytes from Taiwan and the US, with the profile header, the description and the counters in the HTML; rendering adds 78 words for 765,843 bytes. From Japan the same page is 103 words plain and 137 rendered, because the interface follows the exit. A single plurk page is server-rendered at 17,858 bytes. The API needs a signed OAuth request, even for public profiles. On Plurk the setting is plain HTTP, a Taiwanese exit, and an app key for timelines.

Read →
Social proxiesTruth Social Proxies: 1,440 Bytes, No Login

Truth Social measured through residential exits in the United States, Germany and the United Kingdom on 28 September 2026. The public profile page hands a plain HTTP client 7 words from every exit, and a browser hydrates it only from the US: 400 words for 937,369 bytes. The Mastodon-compatible API needs no account and returned the identical 1,440-byte JSON from all three countries, with a documented ceiling of 300 requests per five-minute window. On Truth Social the setting is the API, plain HTTP, any exit.

Read →
Social proxiesVero Proxies: 2,393 Words From 3 Countries

One public Vero profile measured through residential exits in the United States, Germany and the United Kingdom on 28 September 2026. Over plain HTTP it returns 2,393 words in 265,121 bytes from all three countries, identical; rendered, 2,421 words for 331,650 bytes. Every response carries a meta robots tag of noindex, noimageindex, noarchive. There is no public API and no oEmbed endpoint. On Vero the setting is plain HTTP through any residential exit, and the honest line is that the platform publishes profiles to the open web while asking that they not be indexed or archived.

Read →