What Teoma was, and where its results came from
Teoma was an American web search engine that ran its own crawler and its own index — the engine's stored copy of the web, built by software that follows links and fetches pages. It operated under its own brand from a public beta in April 2001 until February 2006. That independence is why it matters. In the early 2000s only a short list of Western engines actually crawled and ranked the web themselves: Google, Inktomi, AlltheWeb/FAST, AltaVista and Teoma. Almost everything else was re-serving results bought from one of them.
The verdict at the top of this page describes Teoma today, and it is unambiguous. On 26 February 2006 teoma.com stopped being a search engine and began redirecting to search.ask.com. A stripped-down version of the name reappeared in mid-April 2010 as what its owner called a simplified interface serving results similar to Ask.com's — a front end, not an index. As of 19 August 2026 the domain does not resolve in DNS at all. There is no Teoma to use, and any list that presents it as a current alternative to Google is two decades out of date.
The engine's working life is a different question, and a more interesting one. For roughly five years Teoma was a real independent index with a ranking idea that nobody else was shipping.
Subject-specific popularity: how it differed from PageRank
Teoma's pitch was a single technical claim: that a link should only count as a vote when it comes from a page that knows something about the subject.
Google's PageRank scores a page by its global link popularity. Every link on the web feeds one enormous calculation, computed ahead of time, and a page's authority is a property of the whole graph rather than of any particular topic. The weakness is easy to state. A wildly popular page about consumer electronics can carry more link weight than a specialist page about plate tectonics, so its outbound links can push an off-topic result up a geology query.
Teoma called its answer Subject-Specific Popularity, and it worked at query time rather than in advance. When a query arrived, the engine assembled the set of pages relevant to it, detected the link community or communities those pages belonged to, and then computed authority scores using in-links from inside that community only. A link from a recognised authority on plate tectonics counted for a geology query; a link from a high-authority but topically unrelated page did not.
This is not PageRank with a filter bolted on. It descends from a different branch of link-analysis research — the hubs-and-authorities line associated with HITS, in which some pages are valuable because they point at good material and others because they are good material. Same family of mathematics, different premise: PageRank asks who the web as a whole respects, Teoma asked who this corner of the web respects. The underlying research began in 1998 at Rutgers University as a project called DiscoWeb, on link-structure analysis of the web.
The cost of the idea was in the same sentence as its appeal. Detecting a topical community per query is expensive computing, done while the user waits, and it does not amortise the way a pre-computed graph score does. That is a large part of why the approach was hard to scale against an opponent doing most of its ranking work offline.
Results, Refine and Resources - a search page built out of the algorithm
Teoma's results page was unusually structured for 2002, and its three panes were not decoration. Each was a byproduct of the same community detection that produced the ranking.
- Results — the web pages themselves, ranked by topic-restricted authority.
- Refine — suggested narrower topics, derived from the labels of neighbouring sub-communities the engine had identified around the query.
- Resources — link lists compiled by identified topic experts: the pages that functioned as hubs, in other words, rather than as authorities.
The design point is worth stating plainly because modern engines still mostly do not do it. Teoma treated an ambiguous query as a signal that the user needed to choose a direction, and it showed the structure it had found rather than flattening everything into ten blue links. Exploratory search of that kind remains badly served today.
How Teoma became the engine behind Ask Jeeves
Ask Jeeves, Inc. acquired Teoma on 18 September 2001, before the engine had even left beta. The reported price is contested and both figures are given here rather than averaged: Wikipedia records "over $1.5 million"; Information Today's contemporaneous NewsBreaks report gives "$1.67 million in cash and stock options". Both are in the same range, both are startlingly small for an engine of this significance, and contemporaneous coverage described the deal as heavily stock-weighted.
Ask Jeeves had begun life as a natural-language question-answering front end over an editorially built knowledge base, with web results supplied by third parties. Buying Teoma gave it an index of its own. The integration date is itself slightly contested: Wikipedia dates it to January 2002, while the Wikipedia biography of Teoma's chief scientist states that Teoma had been the back-end search engine for Ask.com since December 2001. A one-month discrepancy, noted rather than reconciled.
Teoma began as the fallback for questions the Jeeves knowledge base could not answer and became the primary web index. The technology outlived the brand: by December 2010 Ask.com was describing the same lineage as its ExpertRank algorithm. ExpertRank was not a from-scratch replacement invented at Ask — it is the later name for the Teoma line of work.
Who built it, and how it made money
Teoma came out of Rutgers University. Professor Apostolos Gerasoulis, who had led the 1998 DiscoWeb research, founded the company with colleagues in April 2000 to commercialise it; Tao Yang, a parallel-computing academic at UC Santa Barbara, was chief scientist and built much of the crawling and serving infrastructure. Research work sat in New Brunswick and Piscataway, New Jersey; commercial operations ran from Ask Jeeves' base in Emeryville, California, after the acquisition.
The money story is short. Before September 2001 Teoma was a venture-stage company with no meaningful revenue. After the acquisition it was not a business at all but an internal technology asset: revenue was earned at the Ask Jeeves and later Ask.com level, through paid search listings sold or syndicated by third-party ad networks, and Teoma results were also supplied to some partners under OEM-style arrangements. No revenue, funding or contract figures are asserted here because none could be sourced.
One growth claim circulates and needs handling carefully. Citing Nielsen//NetRatings, Wikipedia states that Teoma grew 175% in 2002, making it the third most popular search engine in the United States. Growth-rate rankings from a small base are easy to misread, the underlying Nielsen release is not directly accessible, and the figure almost certainly reflects Ask Jeeves-driven traffic rather than standalone use of teoma.com. No sourced market-share percentage for teoma.com as a destination site was found at all.
What it was bad at
Contemporaneous reviewers were blunt, and their criticisms hold up. Greg Notess catalogued missing basics that competitors already had: no advanced search operators worth the name, no cached copies of pages, no PDF indexing. Lou Rosenfeld questioned the premise itself, doubting that automated clustering could work across the whole web when clustering had historically depended on small, human-curated collections. Both criticisms appear in Information Today's April 2002 report on the engine going live.
Two structural problems compounded them. Index coverage lagged Google's badly, and coverage is the one thing a clever ranking method cannot compensate for — you cannot rank a page you have not fetched. And per-query community detection is costly, which limited how much of the web the approach could be run against economically. Teoma had the better idea about which links should count and the worse position on how many pages it had to count links between.
Its privacy record is not a criticism so much as a blank. Teoma predates the era of published search-privacy policies, no privacy policy from teoma.com was located, and what it logged is simply unknown. The AskEraser search-history deletion feature is sometimes attached to Teoma in retellings; it was an Ask.com feature launched in December 2007, after the Teoma brand had already been retired, and it should not be attributed to Teoma.
How it ended
IAC/InterActiveCorp acquired Ask Jeeves in 2005, taking ownership of the technology. On 26 February 2006 the Teoma brand was retired and teoma.com began redirecting to search.ask.com. In mid-April 2010 the name returned as a minimal front end, with IAC telling Search Engine Land that Teoma.com's organic results "will be similar to, but not exactly the same as, those on Ask.com, primarily for testing purposes". That is a user interface over another engine's results, and it did not last.
Gerasoulis left Ask.com in 2010. Over the following years Ask.com progressively abandoned independent web crawling in favour of resold results, and the Teoma lineage of independent indexing ended with it. A DNS lookup on 19 August 2026 returned no address for teoma.com.
The name, incidentally, is neither an acronym nor a coinage. Teòma is Scottish Gaelic for "expert".
Common misconceptions
- "Teoma was a metasearch engine." No. It ran its own crawler over the open web and built its own index, which is precisely what made it notable in a field full of resellers.
- "Ask Jeeves always had its own index." No. It started as a natural-language front end over a hand-built knowledge base plus third-party results, acquired an index by buying Teoma in 2001, and later gave that independence up again.
- "Teoma's algorithm was just PageRank." No. It descends from the hubs-and-authorities line of link analysis and restricts the vote to the query's topical community. Related mathematics, different premise.
- "ExpertRank was a new algorithm Ask invented." ExpertRank is the later name for the Teoma lineage, not a replacement built from scratch.
- "Teoma is back — I saw a page." The 2010 revival was a cosmetic interface over Ask.com results, described by its owner as a test, and it is long gone.
- "Its expert-community language means it was early AI." It was statistical link analysis. The vocabulary of experts and communities described a graph calculation, not machine learning in the modern sense.
Frequently asked questions
Is Teoma still available?
No. Teoma stopped operating as a search engine on 26 February 2006, when teoma.com began redirecting to search.ask.com. The name was briefly revived in mid-April 2010 as a simplified front end over Ask.com's results, which was described by its owner as a test and did not last. A DNS lookup on 19 August 2026 returned no address for teoma.com at all, so there is no site to visit.
What was Teoma's Subject-Specific Popularity?
It was Teoma's ranking method: instead of scoring a page by its link popularity across the entire web, the engine worked out which topical community of pages a query belonged to and counted only the links from inside that community. The claim was that a link from a recognised authority on a subject should count for queries about that subject, and a link from an unrelated popular page should not.
How was Teoma different from Google's PageRank?
PageRank is global and pre-computed: one calculation over the whole link graph, giving each page an authority score independent of any query. Teoma's ranking was topic-restricted and computed at query time, drawing on the hubs-and-authorities branch of link-analysis research rather than PageRank's. The trade-off was cost. Detecting communities per query is expensive while the user waits, and Teoma's index coverage never approached Google's.
Did Ask Jeeves use Teoma?
Yes. Ask Jeeves, Inc. acquired Teoma on 18 September 2001 and used it as the web-search engine behind its natural-language front end, initially for questions its editorial knowledge base could not answer and then as the primary index. Wikipedia dates the integration to January 2002; the biography of Teoma's chief scientist says December 2001. By December 2010 Ask.com called the descendant technology ExpertRank.
How much did Ask Jeeves pay for Teoma?
Sources disagree and both figures are recorded here. Wikipedia says over $1.5 million; Information Today's contemporaneous report says $1.67 million in cash and stock options. Both are unusually low for an engine of this significance, and contemporaneous coverage described the transaction as heavily stock-weighted. No single authoritative figure has been located, so no single number should be quoted.
Who founded Teoma?
Professor Apostolos Gerasoulis founded Teoma in April 2000 with colleagues at Rutgers University, commercialising DiscoWeb, a 1998 Rutgers research project on link-structure analysis of the web. Tao Yang of UC Santa Barbara served as chief scientist and built much of the crawling and serving infrastructure. Gerasoulis went on to lead search technology at Ask.com and left in 2010.
What does the name Teoma mean?
It is Scottish Gaelic for "expert". The name is neither an acronym nor an invented word, and it was chosen to signal the engine's central idea: that authority on a subject should be judged by the pages that are themselves expert in that subject, rather than by the web at large.
Was Teoma an AI search engine?
No. Teoma was retired years before the large-language-model era and never had AI answers, summaries or chat. Its talk of expert communities described statistical link analysis over a graph of pages, which is a very different thing from the generative systems that carry the AI label today.