Presearch

By Bill Hartzer · Updated 19 August 2026

What Presearch is

Presearch is a crypto-incentivised search engine. You search, and you are paid in PRE, an ERC-20 token, for doing it. Behind the search box sits a network of volunteer-run nodes — ordinary computers running Presearch software — which take your query, fetch results from a range of search engines and APIs, and send them back; node operators are paid in PRE for the work. Advertisers do not bid cash for keywords, they stake PRE against them, and the largest stake takes the slot.

The whole system is a token economy wrapped around a search box. That is the product's real identity, and any description leading with "privacy-focused search engine" rather than "token-incentivised decentralised metasearch" describes the marketing rather than the thing.

The verdict at the top reads resold results, and this is the one page on the site where the verdict is contested by the company itself. Presearch has said publicly since September 2025 that it is building and running an independent index. What follows gives that claim in the company's own terms, alongside the evidence that is missing.

How it gets its results, and what "decentralised" means here

The documented mechanism is decentralised metasearch. Presearch's node documentation is explicit: nodes send the query out to a range of search engines and APIs that return results, which are combined into the results page. That page lists seven possible node operations, one of them Crawling, then says the first release would cover only registering, validating and federating. Federating is metasearch; crawling was aspirational.

Which upstream engines? Presearch's documentation does not name them. Wikipedia lists Google, Bing and DuckDuckGo among those queryable through the network, but the company has never published a definitive provider list. For a product whose pitch is openness, that is a real transparency gap: a user cannot tell whose results they are reading.

The word decentralised needs a caveat too. Nodes distribute retrieval, not authority. Presearch operates the front end, the accounts, the token contracts, the ad marketplace and the provider relationships; turn off presearch.com and Presearch is off. The extra hop has a cost, too — a 2021 academic review noted the decentralised approach may face limitations against centralised systems in indexing and query speed.

The own-index claim, and what is missing from it

Here is the company's position, with dates.

22 September 2025 — the Presearch 3.0 announcement says Presearch is building the first independent web index optimised for "Frontier Intelligence," targeting creator platforms, independent journalism and human-generated content filtered from AI spam, with the system "already crawling and organizing content in stealth." No index size given (press release).
9 December 2025 — company-sourced coverage states that Presearch has completed its transition from metasearch engine to fully autonomous search platform with its own index, describing an AI-native architecture with semantic clustering and nodes that crawl, index and rank.

Here is what is not there. A crawler that visits websites has to be identifiable, so real ones publish a standard set of artefacts: a bot information page, a user-agent string, a robots.txt token, crawl IP ranges, a contact for complaints, an opt-out, an index-size figure. As of 19 August 2026, none of the seven has been published for Presearch. For comparison, Marginalia is a one-person project on roughly $200 a month and publishes all seven.

There is also an internal inconsistency, worth noting without over-reading it: Presearch's live privacy policy still records which provider was used for each search — a field that exists because results come from providers.

Stated as a judgement rather than a fact: the announcements may well be accurate and the index real and growing. On published evidence, though, Presearch is a metasearch engine with an unverified index programme, and this page says so rather than repeating a press release. If a crawler user-agent, IP range and opt-out are published, the verdict should be revisited.

Who owns it now, and how it makes money

Most write-ups still describe the 2017 Canadian project, and that is no longer what Presearch is. It was founded in 2017 by Colin Pape under Presearch.org, a Canadian organisation. Current management acquired the assets in August 2023 and took full control by November 2024; the company now operates from San Diego, California, with Tim Enneking as chief executive. The jurisdiction shift matters for a privacy-marketed product: it moves toward the country most privacy engines advertise moving away from.

Revenue has three documented legs. Keyword staking is the novel one: advertisers stake PRE tokens against specific words and phrases, the highest stake wins the slot, unmatched phrases are decomposed into individual terms, and a recurring consumption fee on the staked amount buys premium placement, with advertisers who decline it still appearing at the bottom of the page. It is a stake-weighted auction with a holding rent, not pay-per-click. Takeover Advertising followed in late 2024 and Spicy Mode in early 2025, opening advertising to adult-content creators — a deliberate choice, since mainstream networks refuse that category. PreGPT subscriptions are the third.

Token economics are a quieter fourth leg: the 2017 crowdsale allocated one billion tokens, and search rewards are paid from its platform-rewards pool, so payouts are subsidised by issuance rather than funded by search revenue. Outside capital is being raised for the index itself — a Wefunder campaign launched March 2025 and live as of 22 June 2025 raised $1,235,000 on a SAFE at a $15,000,000 valuation. The index is the thing being raised for. No revenue figures have been published.

Privacy, and why you cannot be paid anonymously

The privacy policy is more mixed than the marketing. It says clearly that Presearch does not store your search terms. But it also discloses that it stores search events — when a search happened, how many tokens were earned, which provider was used — and collects your IP address for account customisation and fraud prevention. Node operators receive keywords and, where local settings are enabled, general location; IP and browser details are stated as never shared with nodes. No retention period is given; deletion happens on request (privacy policy).

The structural problem the marketing does not address is this: rewards require identity. To be paid in PRE you need an account and a wallet. An account, an IP address, per-search timestamps and token-earning records together form a persistent, linkable behavioural log tied to a financial identity — even with no query text stored at all. That is a different posture from DuckDuckGo, Startpage or Marginalia, none of which need to know who you are. You cannot be paid anonymously; the trade-off is intrinsic to the design.

Two further points. PRE payouts sit on public blockchains, so the existence, timing and size of your reward transactions are durable public data even if your queries are not. And queries transit tens of thousands of volunteer nodes in unknown jurisdictions — withholding identifiers is a real mitigation, but the keyword text still reaches a stranger's machine. No third-party audit was located.

PreGPT, and the absence of AI on the results page

Presearch sells a subscription chatbot called PreGPT, currently version 2.0, in two monthly tiers payable in PRE and listed in its documentation at low single-digit dollars a month. The basic tier runs Mistral 7B, an open-weight model, on Salad.com's decentralised GPU network; the higher tier runs Venice.ai's models through their API. No chat content is stored by default, and conversations are not used for training by default.

Two things deserve flagging. PreGPT is an integration, not Presearch's own model — third-party open-weight models on third-party compute, and the higher tier's "uncensored" characteristic is Venice.ai's property, not Presearch's. And, to the company's credit, PreGPT is a separate paid chatbot rather than an AI answer layer welded onto the results page, which makes the usual question of switching AI summaries off moot here. The "Frontier Intelligence" branding around the index is about indexing human-created content for AI systems, not generating AI answers; the two are frequently conflated.

History and the numbers

2017 — founded by Colin Pape, developed by Presearch.org in Canada, with a one-billion-token crowdsale.
2019 — TechCrunch reports one million users.
Sept 2020 — node beta; Q4 2020 — production launch, limited to registering, validating and federating; crawling deferred.
2021 — added as a default search option on Android in the EU under the European Commission's choice-screen remedy.
Aug 2023 – Nov 2024 — current management acquires the assets, then takes full control; the company becomes US-based in San Diego. Takeover Advertising launches.
Early 2025 — Spicy Mode launches, permitting adult-content advertising.
Mar–Jun 2025 — Wefunder raise for the decentralised web index.
22 Sept 2025 — Presearch 3.0: a Base-native Web3 rebuild with gas-efficient microtransaction payouts, 40,000-plus nodes claimed, the index said to be crawling in stealth.
9 Dec 2025 — the company announces the transition from metasearch to an independent index is complete.
2026 — still operating. No lawsuits, enforcement actions, breaches or delistings located.

Every usage figure is company-supplied and none independently audited. As of December 2025 the company gave over 400,000 daily searches and over 12 million a month — internally consistent, so one claim rather than two — plus 150,000 active monthly users, and 40,000-plus nodes as of September 2025. Two figures are in tension and this page does not reconcile them: TechCrunch reported one million users in 2019 against 150,000 monthly actives in December 2025. Either the 2019 number counted registrations, or usage fell substantially. There is no market-share figure — Presearch is not broken out by StatCounter, and 12 million monthly searches is roughly what Google handles in a few seconds.

Who it suits, and who it does not

What is genuinely interesting. Presearch is the only serious attempt to attach an economic incentive to the demand side of search: nine years running, an ownership change survived, a working node network shipped, a place on the EU Android choice screen. Keyword staking is a mechanism nobody else runs, and no mandatory AI summary has been bolted onto the results page.

What it is bad at is the two things it markets hardest. Independence: the documented mechanism is fetching other engines' results, and the claimed own index has produced no crawler, no user-agent, no index size and no webmaster-facing documentation after a year. Privacy: the reward model requires an account, an IP address, a wallet and a per-search event log, and payments are permanently public on-chain. It also will not tell you whose results you are reading. Result quality is inherited, the extra hop costs speed, and the token adds a failure mode no other engine here has: PRE's price is volatile and rewards come from issuance rather than search revenue.

It does not suit anyone with a real privacy threat model, for whom an account-linked search-event log plus an IP plus a public payment trail is worse than an anonymous session elsewhere; or anyone choosing it to escape Google and Bing, because on published evidence those results are what you are reading, routed through a stranger's computer first.

Frequently asked questions

Does Presearch have its own index?

The company says yes; the evidence has not been published. Presearch announced an independent index in September 2025 and said in December 2025 that the transition from metasearch was complete. But no crawler name, user-agent string, robots.txt token, crawl IP range, opt-out mechanism, webmaster contact or index-size figure has appeared. Its own node documentation describes fetching results from other search engines and APIs.

Where do Presearch results come from?

The documented mechanism routes your query through volunteer-run nodes that fetch results from other search engines and APIs and combine them into one page. Presearch has never published a definitive list of which providers those are. Wikipedia names Google, Bing and DuckDuckGo among the engines that can be queried through the network. Its privacy policy still logs which provider served each search.

Is Presearch really decentralised?

Partly. Tens of thousands of volunteer-run nodes distribute the work of fetching results, which is real. But the company in San Diego operates the front end, the accounts, the token contracts, the advertising marketplace and the provider relationships. Nodes decentralise retrieval, not authority: if presearch.com goes dark, Presearch goes with it. It is crowdsourced labour rather than an ownerless network.

Can you actually earn money using Presearch?

You earn PRE, an ERC-20 token, at fractions of a cent per search, paid out of a platform-rewards allocation created in the original 2017 token sale rather than out of search revenue. The token's price is volatile. Treat any framing of Presearch as an earnings opportunity with scepticism: the amounts are small and the asset can move sharply.

Is Presearch private?

Less than the marketing suggests, and for a structural reason. Search terms are not stored, but search events are — timestamp, tokens earned, and which provider served the query — and your IP address is collected. Being paid requires an account and a wallet, so an identity is attached by design, and PRE payouts are permanently visible on public blockchains. You cannot be paid anonymously.

Who owns Presearch now?

Not the people who founded it. Colin Pape founded Presearch in 2017 under Presearch.org, a Canadian organisation. Current management acquired the assets in August 2023 and took full control by November 2024, and the company now operates from San Diego, California, with Tim Enneking as chief executive. Most third-party write-ups still describe the original Canadian project.

What is PreGPT?

A subscription chatbot sold separately from search, in two monthly tiers payable in PRE. The basic tier runs Mistral 7B on Salad.com's decentralised GPU network; the higher tier runs Venice.ai models through their API. Both are third-party open-weight models on third-party compute rather than Presearch's own. Chat content is not stored by default and is not used for training by default.

How many people use Presearch?

Company figures only, and two of them conflict. TechCrunch reported one million users in 2019; the company reported 150,000 active monthly users and over 12 million searches a month in December 2025. Both are recorded here without reconciliation. No independent audit exists, and Presearch is not broken out in StatCounter data, so its market share is unmeasured.