Original research · 24 September 2026
22 years of Wikipedia debates
Wikipedia’s editors spent years arguing over whether search engine optimization was a profession or a kind of spam. We coded 1,288 edits from those debates, and from the first edits of the two articles at their center, to measure when the argument turned.
What this research found
On the talk pages where Wikipedia’s editors argued about SEO, the share of comments framing it as spam fell from 40% (10 of 25 that took a side) in 2004-05 to 22% (23 of 103) in 2006-07, as legitimate-profession framing surged: 14 comments in 2004-05, 74 in 2006-07.
Two AI coders (Claude Sonnet models), working independently from a written codebook, labeled every contribution spam, legitimate, mixed or neutral; a contribution took a side unless it was neutral (see the method). The counts are distinct contributions, the first appearance of each text on its page, so a comment that was re-saved, restored or corrected counts once. All 25 from 2004-05 are on the SEO talk page. In 2006-07, 99 of the 103 are there and 4 are on the Spamdexing talk page.
| Years | Took a side | Spam | Legitimate | Mixed | Spam share | |
|---|---|---|---|---|---|---|
| 2002-03 (article edits) | 5 | 4 | 1 | 0 | 4 of 5 | |
| 2004-05 | 25 | 10 | 14 | 1 | 40.0% | |
| 2006-07 | 103 | 23 | 74 | 6 | 22.3% | |
| 2008-12 | 33 | 7 | 26 | 0 | 21.2% | |
| 2013-26 | 2 | 1 | 1 | 0 | 1 of 2 |
2002-03 holds only the first edits of the Spamdexing and Search engine optimization articles: the SEO talk page’s first revision is dated 13 May 2004. The coding covers the SEO talk page through 2012, so from 2013 it covers only the Spamdexing and Keyword stuffing talk pages. Where fewer than 20 contributions took a side, the table gives the count, not a share.
The share then held at about a fifth: 21% (7 of 33) from 2008 through 2012. After that, 2 contributions to the pages coded took a side.
2002
On 10 March 2002, an anonymous editor created the article Spamdexing and defined it as the practice of “deliberately modifying HTML pages to increase the chance of them placing high” on a search engine’s relevancy list.
Three months later, on , Tarquin rewrote the opening to explain the word: “a portmanteau of ‘spam’ and ‘indexing’”. The encyclopedia had a name for gaming search results eleven and a half months before it had an article on the work itself, a gap our study History of SEO Terms on Wikipedia measures.
2003
Sherwoodseo created Search engine optimization on . The term, it said, also refers to “an industry of consultants that carry out optimization projects on behalf of client sites”.
Thirty-six days later, on , LittleDan added that SEO is often confused with spamdexing, “the illegal promotion of pages through taking advantage of the search algorithms”. The next day an anonymous editor changed one word, with the edit summary “illegal -> unwanted”: “the unwanted promotion of pages”. On another anonymous editor added that many search engine administrators say “any form of search engine optimization used to improve a website’s page rank is nothing else than spamdexing”.
Counted as distinct contributions, the founding edits of the two articles hold 5 that took a side, and 4 of them framed the practice as spam: the Spamdexing article’s creation, its etymology, LittleDan’s sentence and the claim of 7 April 2003. The other, the SEO article’s own first revision, framed SEO as a legitimate industry. The one-word softening does not count on its own: it corrects a sentence that was already there.
2004
The SEO article’s talk page opens on 13 May 2004, more than a year after the article. Kim Bruning started it by apologizing for removing all of the article’s external links: “fwapping SEOs all day, and I’m losing my patience”.
Jmabel answered that night that “there is ethical and unethical SEO”. The next afternoon Kim Bruning warned that an SEO’s tutorial can turn out to be an ad: “they’re advertising anyway”.
On 17 May 2004 the argument turned to one link. An anonymous editor objected to its removal and said they would restore it: the page, they wrote, was initially part of a search engine optimization competition. The next day, after a second objection, Kim Bruning wrote: “That page was link-pushing Nigritude Ultramarine, at the freaking top of the page even”. An anonymous reply: “‘Link-pushing’ as you describe it, is a tactic used in search engine optimization”.
In 2004-05, 10 of the 25 distinct comments that took a side framed SEO as spam, 40%.
2006-07
On the account Dannysullivan answered a warning from Perfecto about adding commercial links. The note, left on Perfecto’s talk page and signed “Danny Sullivan, Editor, Search Engine Watch”, argued that SEO does not include paid advertising: “SEO is akin to PR, optimizing for non-paid improvement”. Four days later Perfecto copied it to the SEO talk page: “He sent me this note on my Talk page”.
The links were still the fight. Synlighet argued for keeping the article free of them: “Every SEO SEM company, forum, guru, professional would love to have a backlink from this article”. On Rasmus Faber moved a passage out of the article and onto the talk page: “It reads like an advertisement for a SEO-company”.
The same years changed the vocabulary. In 2004-05, 11 distinct contributions used the words ethical or unethical and 2 spoke of white, black or gray hats. In 2006-07 the hats led, 23 to 8, and they still led from 2008 through 2012, 11 to 3. On the Spamdexing talk page, Ethicalmarvin wrote on : “‘Spamdexing’, at best, is a subset of ‘Black-hat SEO’”.
| Years | Contributions | Ethical or unethical | White, black or gray hat |
|---|---|---|---|
| 2002-03 | 7 | 0 | 0 |
| 2004-05 | 117 | 11 | 2 |
| 2006-07 | 455 | 8 | 23 |
| 2008-12 | 305 | 3 | 11 |
| 2013-26 | 32 | 0 | 2 |
On an anonymous editor removed the {{Spamming}} template from the article and explained on the talk page: “Spamming template is derogatory to SEO business”. ZimZalaBim replied minutes later: “The template should remain.” Jehochman called the objection valid, suggested the template cover only the article’s black hat material, and added: “There is a POV that all SEO is spam.” Wikipedia, the comment went on, should not take sides.
On 7 March 2007 an administrator semi-protected the article, giving the reason “Spam magnet”. On 1 June 2007 it was promoted to Featured Article, and on 2 July 2007 it ran on Wikipedia’s Main Page. That day Redquark objected that the article played down black hat SEO: “in my circles, ‘SEO’ is often a shorthand for spammy tactics”, and reported editing the lead to mention spamdexing.
In 2006-07, 23 of the 103 distinct comments that took a side framed SEO as spam, 22%; 74 framed it as legitimate.
2008-12
In September 2010 Texxs added one line to a thread opened in 2007 under the heading “Notable SEOs not acceptable”, about the practitioners the article named as notable: “some of them are obvious shills for Google”.
In September 2011 Cantaloupe2 marked the article’s citations of an SEO industry publication as dubious, after finding “an article it published on how to spam Wikipedia and stay under radar on its site”. In the dispute that followed, Jehochman asked: “What sort of bias are you alleging me to have?” and went on: “My bias is no different than that of any published expert”.
On 3 April 2012 an administrator semi-protected the article indefinitely for “Persistent spamming”, and on 14 October 2012 it lost its Featured Article status after a review. From 2008 through 2012, 7 of the 33 distinct comments that took a side framed SEO as spam, 21%.
2006 to 2024
The article on gaming search engines is still called Spamdexing. Editors tried to rename it, turned the word down for a category, and finally merged another article into it.
On LinguistAtLarge opened a thread on the Spamdexing talk page: “I vote that we move this page to Black Hat SEO”. SteveBaker added that November, of the word: “It’s a neologism and it fails the Google test”. The last comment came from Philwiki on : “I rarely hear of the word spamdexing”. The thread was archived in 2016 without a decision, and the article kept its name.
On Jehochman proposed renaming Category:Black hat seo to Category:Spamdexing: “The most correct term for black hat seo is Spamdexing”. Choster disagreed: “‘spamdexing’ is a relative neologism”. Johnbod agreed with that: “I’ve been aware of this for years, but had never heard of ‘spamdexing’”. The proposer conceded: “I see nothing wrong with Category:Black hat search engine optimization”. On 9 January 2009 Kbdank71 closed the discussion with the result: rename to Category:Black hat search engine optimization.
The last debate ran the other way. On BobsFullBogs wrote on the Keyword stuffing talk page: “I propose we merge this page with Spamdexing”. Gilgongo agreed on : “there is too much overlap”. On Klbrain closed it, “Given that there are no objections”, and merged Keyword stuffing into Spamdexing.
What professionalization changed
In 26 comments, an editor said in their own words that they did SEO work. Of those 26 comments, 13 framed SEO as legitimate, 13 took no side, and none framed it as spam.
That finding is about those comments, not about the people who wrote them. They came from 13 registered accounts and 7 IP addresses. On the coded pages, the 13 accounts made 205 distinct contributions, and 4 of them, from 3 of the accounts, framed SEO as spam.
Two of the comments: NetOracle, “I am an SEO professional”; and Bill Slawski, “SEO is marketing, and not an IT department function”.
Coders also tagged each contribution with its themes. Among the 168 distinct contributions that took a side:
| Theme | Contributions | |
|---|---|---|
| SEO as a business or profession | 130 | |
| External links, advertising and self-promotion | 89 | |
| Ethics, white hat and black hat | 77 | |
| Neutrality and fairness to the industry | 57 | |
| Search engines' rules and penalties | 43 | |
| What things should be called | 37 |
Spamdexing is dated in our study History of SEO Terms on Wikipedia. The other terms in this debate are in the SEO glossary: SEO, Black hat SEO, White hat SEO, Keyword stuffing, Link spam.
On 24 September 2026 we made 70 read-only requests to Wikipedia’s API and fetched revisions, with the text of each, from 19 pages: the full histories of Talk:Search engine optimization (2,184 revisions) and its 10 subpages, which are 7 archives, a redirect, a draft and a to-do list (125); Talk:Spamdexing (90) and its 3 archive pages (5); Talk:Keyword stuffing (40); the Categories for discussion log of 3 January 2009 (227); and the first revisions of the Spamdexing (6) and Search engine optimization (17) articles. That is 2,694 revisions. Each was compared with its parent, and the text it added was credited to its editor. Leaving out the Categories for discussion log, which we quote directly, and edits that only removed text, or added only whitespace, a bare signature or a bot’s archiving, that gives 2,081 records.
1,288 records were coded: both articles’ first edits, the SEO talk page from 2004 through 2012, and the Spamdexing and Keyword stuffing talk pages in full. The coders were AI models. Two Claude Sonnet coders, working independently from a written codebook, coded every record: the framing (spam, legitimate, mixed or neutral), the themes, whether the editor said they did SEO work, and a verbatim quote of 20 words or fewer. They agreed on the framing of 89.3% of records (Cohen’s kappa 0.735). A third model, Claude Opus, settled the 159 records where they disagreed on framing (138) or on self-identification (31), under 6 written rules. Claude, an AI assistant, drafted the codebook and its rules. Kyle Alm approved the outline and chose the headline measure from the recounted results.
Each coder saw at most the first 2,000 characters of a record’s added text. 93 of the 1,288 coded records were longer, 50 of them distinct contributions. 27 of those distinct contributions are from 2006-07: 20 coded neutral and 7 legitimate. The vocabulary counts read the full text.
The text a revision adds is whole lines, so a comment is coded again whenever someone re-saves, restores, moves or corrects it. We count each text once, at its first appearance on its page. A record is a repeat when the words it inserted mostly (more than half) repeat text already on that page in an earlier revision, or when its coded quote was already there. The inserted words are measured word by word, the way Wikipedia’s own diff view shows them, so a sentence appended to an existing paragraph counts as new; a change of a few words is judged by the line it sits in. The quote test is approximate, on five-word sequences, so a quote that differs from earlier text only by a corrected typo counts as already there. A record that inserts no words of its own does not count. 916 of the 1,288 records are distinct contributions. We also ran the recount with sequences of four, five, six or eight words and thresholds of 30%, 50% and 70%: the 2004-05 and 2006-07 shares hold at every setting tested (10 of 25, and 22% to 23%), and 2008-12 stays between 21% and 27%.
Early comments are often unsigned, so authorship comes from each revision’s record of its editor. Credit goes to the editor of the revision that added a text, so text that one editor moves or pastes carries the mover’s name. So every quote on this page was checked to first appear in the diff it links, across all 19 pages fetched; the one note that was copied in from elsewhere was traced back to the user talk page where it was first written. Anonymous editors are not named: an IP address is not a person. The archive pages and three small subpages (a redirect, a draft and a to-do list) were fetched but not coded; the archives repeat the live page’s history. The Categories for discussion log was fetched to be quoted, not coded. The SEO talk page’s first revision is dated 13 May 2004, so the earliest era holds only article edits, and only 5 of them took a side. Counts are of contributions, not of people: a prolific editor can account for many.
The revision histories come from Wikipedia’s own API. Automated tests recompute every count of revisions, records and contributions on this page, and the coders’ agreement, from the committed data; a verifier re-fetches every quoted revision from Wikipedia to confirm the words, the editor and the time. Built with Claude.
Wikipedia text quoted on this page is by Wikipedia contributors, available under the Creative Commons Attribution-ShareAlike 4.0 International license (CC BY-SA 4.0). Every quote links to the revision it comes from.
SEO Bandwagon, "When SEO Was Spam: 22 Years of Wikipedia Debates" (2026-09-24). https://seobandwagon.com/research/when-seo-was-spam
This research came out of the same tooling we point at client sites: archives, revision histories and search data, read as evidence rather than anecdote.
Browse the SEO glossaryor talk to us · History of SEO Terms on Wikipedia · GBP benchmarks research