GEO and AEO on Wikipedia

Part one

Who named it, when, and where it stands

Four questions have different answers for each name. GEO has a paper, a date and six named authors. AEO has an older but vaguer origin, and a claimant whose evidence is a year late. SEO has a position, and it is that both are SEO.

Sources for the off-Wikipedia rows are in Part seven.

Part two

The naming record: 70 events, settled without a debate

The contest over the names was settled by speedy deletions, draftifications and bold redirects, not by argument. 70 dated, signed events across 30 titles, from December 25, 2024, to June 16, 2026: 24 creations, 22 deletions (6 for advertising, 6 for abandoned drafts, the rest housekeeping), 14 page moves, 8 redirects and retargets, and 2 protections.

Data: naming-events-2026-10-31.csv.

Six titles were deleted for advertising, from June 1, 2025, to March 9, 2026. Nine redirects now point at the GEO article: Artificial intelligence optimization, AI optimization, Answer engine optimization (in three spellings), AI SEO, Artificial Intelligence Search Engine Optimization, a draft and a user page. The SEO article has 79.

The wrong inventors, and the round trip

The first English GEO article, live on April 24, 2025 (Wayback capture of revision 1287105237), credited the term to six researchers named Gao, Liu, Si, Meng, Xiong and Lin, beside the correct arXiv number, 2311.09735. arXiv lists the authors as Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande. Hebrew Wikipedia translated that article on May 11, 2025, and Catalan carried the same wording from its first version on May 17, 2025. Both still credit Gao and colleagues today (Hebrew revision of 6 October 2026, Catalan of October 23, 2026).

Today’s English article is a round trip. It was created on September 28, 2025, by translating the Hebrew article, which had itself been translated from the deleted English one. The translator replaced the wrong authors with Aggarwal and colleagues on the way back.

The name arrived as AIO

The earliest title of all is a draft, Generative Engine Optimization (GEO), created on December 25, 2024, and declined the same night as “not notable” and AI-written. The first article to reach and stay in mainspace was Artificial intelligence optimization: drafted January 10, 2025, created 25 February 2025, tagged AI-generated within four hours, draftified the next day, and back in mainspace on May 3, 2025.

The merges carried the names, not the text

Of each absorbed article’s last version, only 8.3% (AIO), 7.4% (AEO) and 1.8% (AI SEO) of its three-word phrases appear in the GEO article now. The GEO article itself keeps 4.8% of the phrases of its own first version. It was stubbed on March 1 2026 for LLM-generated text that failed verification, and two LLM-written expansions of more than 21,000 bytes each were reverted on March 24 and 26, 2026.

The queue behind it

22 Draft pages mention the phrase: 16 about GEO companies and products, 4 about people in the trade, 2 about the concept under another name (AI Search Optimization and AISO). 47 user talk pages carry it. This is the first study’s “then the companies” pattern, arriving as drafts rather than as deletions of established articles.

Other language editions

23 Wikipedias have a GEO article. English was first (April 23 2025) and Bulgarian next (April 24, 2025), but English lost its first two articles, so today’s English article is the 11th of the surviving editions to be created. Romanian files it under the title AEO (August 2, 2026), Spanish as Optimización de modelos de lenguaje (LLM optimization), Estonian as AI visibility optimization.

Part four

Where SEO sits: 20 months before it mentioned any of this

No stable version of Search engine optimization used “generative engine optimization”, “LLM”, “ChatGPT” or “AI Overviews” between November 2022 and June 2025. The first addition that stuck is dated July 21, 2025.

Today the section says digital marketers developed AEO or GEO for LLM-based search, that the terms are used interchangeably, and names one more: “AI engine optimization (AIEO)”. Of the SEO article’s 343 versions since November 30 2022, 87 (25%) were later reverted, close to the GEO article’s 21%.

Part five

Word drift: AIO led through 2025, GEO and AEO run level now

Measured as in the first study: density per 1,000 content words (stop words and tokens of two letters or fewer dropped), here for every live GEO, AEO, AIO and AI SEO article at each month end, pooled. By October 2026 GEO and AEO run nearly level inside the one surviving article, with SEO close behind. That is the Terminology section doing its job: the article names every rival and so carries the debate in its prose.

Data: geo-family-monthly-2026-10-31.csv, page-versions-terms-2026-10-31.csv.

Phrases that faded, arrived and endured

Across the family’s 21 month ends (floor: present in 3 months, 2 occurrences in the peak month). The faded list is the vocabulary of the absorbed AIO and AEO articles: what “a system” does (retrieval, AI-generated content, AI-powered search) and the vendor language of monitoring brand representation. What arrived is the vocabulary of naming (“aeo geo”) and of tools. “structured data”, which the first study found still peaking across the SEO corpus, vanished from the GEO family in 2025.

Data: ngram-discovery-2026-10-31.csv.

In the SEO article

The AI vocabulary that arrived in July 2025 (“artificial intelligence”, “large language models”) had left the prose again by February 2026; “generative engine optimization”, “authority signals”, “user queries” and “digital marketers” stayed. “Internet marketing” fell to 29% of its 2022 density while “digital marketing” arrived in August 2025, the same replacement the first study dated to 2016 across its corpus.

Text survival

The current GEO article retains 4.8% of the three-word phrases of its first version. The largest single rewrites were the stubbing of March 1, 2026, and the reverted expansions of March 24 and 26, 2026 (similarity to the previous version 0.03).

Data: geo-article-text-survival-2026-10-31.csv.

Part six

Who the GEO article cites, version by version

The GEO article’s sources moved from academics and vendors to the trade press. Its first version cited 11 sources: 3 academic, 3 vendor or agency, 3 news. The current version cites 27: 13 search and marketing trade press (Search Engine Land, Search Engine Journal, TechTarget, Search Engine Roundtable, MarTech, Digiday), 5 news, 3 vendor, 2 Google and Bing, 2 analyst. The founding paper is still cited, by arXiv number only.

In the SEO corpus general journalism displaced the industry’s own sources. Here the industry’s own trade press is displacing both academia and vendors, while editors on the talk page question Search Engine Land’s independence after its sale to Semrush.

The article deleted on June 1, 2025, is captured once, by the Wayback Machine on 24 April 2025: seven references and no links, namely the paper (with the wrong authors), Search Engine Land, and five vendors or agencies (Walker Sands, Ahrefs, Peec AI, Otterly AI, Profound). The AIO article at its last live version had 5 external links and 6 identifier citations on 123 content words; a talk thread called it “ref bombing”.

The most contested page

The GEO article has been semi-protected twice for disruptive editing (January 22, 2026 and May 6, 2026). 46 later-reverted edits added 129 URLs, 69 of them (53%) to vendor and agency sites. Its peak was 60 external links on October 26, 2025, 38 of them to vendors and agencies, reverted 41 minutes later as AI-generated text.

Later-reverted versions added 129 URLs to the GEO article, 50 to AIO, 14 to AI SEO, 8 to AEO and 57 to the SEO article since 2022. The domains most often added and removed again on the GEO article: searchengineland.com (5), buzzsense.ai (3), then similarweb.com, peec.ai, writesonic.com, seranking.com and other GEO tools (2 each). Source classes were assigned by hand to all 283 cited domains; an unlisted domain is a vendor, agency or tool site.

Data: page-versions-links-2026-10-31.csv.

Part seven

The record outside Wikipedia

The claims we could read at the primary source hold: the GEO paper and its authors, the 2018 Search Engine Watch article that first calls the practice AEO, and the 2022 Hacker News comment that uses the phrase in the LLM sense. The claims we could not check are marked.

Not found, which is not proof of absence: any use of “generative engine optimization” before November 16, 2023, any use of “answer engine optimization” before 2018, dated first uses of “SGE optimization”, “AI SEO”, “AIO” and “voice search optimization”, and who coined “position zero”.

Part eight

Which name searchers use

Searchers agree with Wikipedia on GEO over AEO, but neither is the name the market searches. The GEO phrase has out-searched the AEO phrase in every quarter since mid-2025, by 1.6 to 2.0 times worldwide and 1.4 to 1.9 times in the US. “ai seo” runs 4.9 to 11.6 times above the GEO phrase and 7.8 to 23.2 times above the AEO phrase, and “ai search optimization”, the name a third of the Search Engine Land survey’s respondents said their clients use, runs level with GEO (0.9 to 1.2 times).

Google Trends, web search, weekly, 2022-11-01 to 2026-10-30. Two worldwide batches share “ai seo” as an anchor, with identical values in every week, so they sit on one scale (100 = “ai seo” in the week of May 3, 2026). “Last 4 weeks” uses the four most recent complete weeks against the term’s highest week.

Data: trends-weekly-2026-10-31.csv, trends-quarterly-2026-10-31.csv.

• AEO registered first. In US Trends the AEO phrase shows a value of 1 in two weeks of September 2023 and April 2024; the GEO phrase is zero until February 2025, worldwide and in the US. That fits the 2018 origin.
• Every name peaked in May to July 2026 and has since fallen. Worldwide the GEO phrase is at 40% of its peak and “ai seo” at 30%; “llm optimization” kept only 12%. In the US every term fell by half or more between the weeks of July 5 and 19, 2026. A drop that size, in every term in the same two weeks, may be a change in how Trends samples rather than in what people search. Google Ads does not show it (below).
• Searchers attach the acronyms to SEO. By Q4 2026 “geo seo” (7.3) out-searches the spelled-out GEO phrase (5.7), and “aeo seo” (4.0) the AEO phrase (3.3).
• “ai seo” is ambiguous. It covers AI tools for SEO as well as SEO for AI answers, the same confusion the Wikipedia editors hit before redirecting AI SEO in March 2026. Read it as an upper bound.

US monthly search volume

Requested but not returned: “ai visibility optimization”, “search everywhere optimization”, “geo vs seo”, “aeo vs seo”, “aeo vs geo”, “llm seo”. Part ten has Labs figures for the comparison phrases and “llm seo”.

Data: search-volume-us-2026-10-31.csv.

The US summer fall is a Trends reading, not a fall in searching

We checked the Trends data against a download from the Trends website on November 1 2026 (United States, past 12 months, web search): the same query through DataForSEO matched it week for week.

The week of 2026-11-01 is partial and excluded. “ai seo” here and in the 2026-10-31 US batch, which had different companion terms, correlate at 0.987 over 52 weeks.

Data: trends-website-vs-dataforseo-us-2026-11-01.csv.

Google Ads does not show the US fall, and the SEO control did not halve. From May to August 2026 the spelled-out names fell 55 to 67% in Trends while Ads has them flat or rising.

Trends: monthly mean of the weekly US index. Google Ads: monthly volumes in the Labs keyword_overview pull of 2026-10-31. The names sit at index 2 to 8, where one point is 12 to 50% of the value. Worldwide has no Ads check.

• The bare acronyms do not measure the disciplines. GEO’s top related queries are “nat geo” (79 to 82), “geo dash” (56), “geo guesser” (50 to 51), “geo group” (48 to 49), “geo tracker” (46), “neo geo” and “geo metro”. So bare GEO’s mean of 28 against AEO’s 3 is not a 9-to-1 lead; the spelled-out phrases put GEO 1.4 to 1.9 times ahead in the US. Bare AEO is mostly the marketing sense, with American Eagle’s ticker, “aeo stock”, at 37 to 49.
• SEO stepped down by a third in the week of 2026-09-27 (56 to 58, then 36 to 37, in all three pulls) and held there for five weeks; the spelled-out names did not move that week. Not explained.
• Automated traffic shows in the related queries. Both sources list everyday searches as rising with AEO: “best pizza near me”, “what is a 401k”, “cheapest flights to tokyo”, “how to bake a cake”, “vegetables” (rising 21,900%) and others. People researching AEO do not search for pizza and Roth IRAs in the same sessions at these rates; automated sessions that mix everyday queries into their searches are the likelier source. That is an interpretation, not a measurement. At index 2 to 8, AEO is the line such traffic moves most.
• Vendors are behind AEO’s rise among searchers. “hubspot aeo” rose 700% (DataForSEO) and 750% (Trends website), and “hubspot” alone is on AEO’s top list at 23. “profound aeo” rose 200% in both. HubSpot’s blog is also the second most-cited source in the AI answers of Part ten (13 of 72).
• Searchers compare the names. AEO’s top related lists hold “seo” (56 to 69), “geo” (46 to 58), “aeo geo”, “aeo vs seo” and “seo geo aeo”. The comparison volumes in Part ten say the same (“geo vs seo”, 2,900 a month).

Part nine

The “King of AEO” press releases

In September 2026, at least three SEOs put out wire press releases announcing they had been “named King of AEO”. Asked on November 1, 2026, who holds the title, every engine we checked named the first claimant first. Google’s AI Overview and Gemini now also call the title a meme or an experiment, not an award. The query registered in US Trends on 3 of 93 days, so it is too small to move the US AEO line.

We do not name the claimants. The three releases are cited by outlet and date, as published, with nofollow links: PRLog, September 1, 2026, ABNewswire, September 7, 2026, ACN Newswire, September 8, 2026. The claimants’ own pages call it an experiment.

One run each on November 1, 2026. AI answers change; read this as a dated observation, not a standing result.

The releases pass no link value; the headline does the work

Every link to a claimant site on the 7 release copies we read is nofollow (8 of 8), and two of the copies (openPR and FinancialContent) carry none. Three copies point their canonical at the syndicator, getnews.info. What an engine takes is the text, and Perplexity cites the copies directly. The pickup is the press-release sections of news sites: Wedbush’s investor site and FinancialContent, WBOC, Daily Tribune, NBC Right Now, StreetInsider, Barchart, USA Today, AZ Central, National Law Review, Yahoo Finance and openPR.

It became a scored contest

At least 7 dated releases ran from September 1, 2026, to October 13, 2026: the three above, a release on September 23 announcing the first claimant’s new “discipline”, two October standings releases (October 10 and 12, per one entrant’s log) and a regional spin-off on October 13. A competition site measured 8 claimants on October 13 across ChatGPT, Gemini, Google AI Overview and AI Mode, Perplexity and Claude in the US, UK and Australia (117 scored answers, 12 Google result sets) and left the title vacant on October 14. One claimant conceded to the first on October 29.

The contestants search the query by script

The competition’s October 13 run and one entrant’s logged pulls (3 runs per engine on the evening of September 12, 10 per engine on September 30) are automated searches for this exact question. US Trends’ one spike is September 12 (100); worldwide peaks on September 10 (100) and September 25 (89); the two late US values, October 29 and 31 (6 and 11), follow the concession. AEO held at 2 in every week from August 23 to September 27 in a separate four-term pull, the spike’s week included. That is a match of dates, not a measurement.

Data: king-of-aeo-trends-daily-2026-11-01.csv.

Contest content now writes about the naming debate itself

A history blog whose newest dated post is from 2015 now carries a page saying the AEO-versus-GEO naming debate began as an exchange between two claimants, and a domain that imitates a newspaper’s history section carries a “royal family of SEO” timeline. A rival’s concession describes the first claimant’s method as placing supporting content on outside sites. An AI answer about who named AEO or GEO could draw on such pages. The origins in Part seven come from primary sources and stand; we do not cite either page.

Sample: 22 pages found by web search, 17 could be read (3 refused by the site, 1 disallowed by its robots.txt, 2 gone). Answers from Google AI Overview, the ChatGPT app, Perplexity and Gemini.

Part ten

What the names cost, and how AI engines search them

AI engines translate the rival names into GEO, and advertisers pay most for the plain-English name. Asked “What is AI SEO?”, ChatGPT searched for “generative engine optimization” in 3 of 3 runs. Asked “What is AI search optimization?”, ChatGPT and Gemini added GEO to all 5 of their searches. On two prompts that named no discipline, 7 of the engines’ 21 searches carried one, and 6 of those 7 said GEO. In paid search, “ai search optimization” has the highest cost per click of the head terms, $48.74, and the AEO phrase the lowest, $21.31.

Cost per click

Short forms: “geo seo” 2,400 at $23.38, “aeo seo” 720 at $20.28, “geo agency” 590 at $41.09, “aeo agency” 590 at $60.01, “geo tools” 720 at $30.09, “aeo tools” 390 at $39.53, “llmo” 480 at $18.67, “llm seo” 720 at $40.33, “ai visibility tools” 880 at $67.53. Comparisons: “geo vs seo” 2,900 at $12.12, “aeo vs seo” 1,900 at $10.58, “aeo vs geo” 1,600 at $13.88, “geo vs aeo” 880 at $8.65. “answer engine optimization tools” shows a CPC of $2.93 with a low top-of-page bid of $17.46; the CPC contradicts its own bid range, so it is not quoted.

• The plain-English name costs most. “ai search optimization” leads at the head ($48.74) and with “agency” ($78.90), above “seo agency” ($48.71). Labs labels it commercial; the GEO and AEO head phrases are labeled informational.
• GEO costs more than AEO at the head, $35.04 against $21.31, and more than plain “seo” ($27.38). With “services” the two are level ($49.08 and $51.67). The acronyms reverse it: “aeo agency” $60.01, “geo agency” $41.09.
• Comparison searches are large and cheap. “geo vs seo” (2,900) and “aeo vs seo” (1,900) out-search every “+ agency” form at $10 to $14 a click: a content opportunity more than an ad market. Searchers put AEO first more often: “aeo vs geo” 1,600, “geo vs aeo” 880.

Data: cpc-us-2026-10-31.csv.

Fan-out queries

A fan-out query is a web search an AI assistant runs while answering one prompt. Twelve prompts went to three engines on October 31, 2026: the ChatGPT consumer app through a scraper, United States, web search forced, 3 runs a prompt; Gemini 3.8 Flash through its API with search grounding, 2 runs; and Claude Sonnet 5.5 through its API with web search, 1 run. Google AI Mode and Perplexity were probed and return no fan-out field.

Data: fanout-queries-2026-10-31.csv, fanout-answers-2026-10-31.csv.

• The engines fold rival names into GEO, never into AEO. ChatGPT expanded “AI SEO” to “generative engine optimization” in all 3 runs; Gemini kept “AI SEO”. For “AI search optimization” ChatGPT and Gemini added GEO in all 5 of their searches; Claude searched the phrase as given. AEO entered a search only where the prompt named it or alongside GEO.
• With no name given, GEO is the one the engines choose: 6 of the 7 searches that carried a discipline name, out of 21 on the two neutral prompts. ChatGPT mostly sidestepped the naming question: 11 of its 15 neutral searches went to platform documentation (OpenAI’s OAI-SearchBot guidance, Google Search Central, Bing Webmaster Guidelines).
• “LLM optimization” is not yet a marketing name to the engines. All three answered “What is LLM optimization?” with model engineering first (quantization, inference, training). ChatGPT’s 3 searches were all in that sense, and Gemini mentioned SEO in neither run. The 720 US searches a month mix both senses.
• Answers name GEO more than any rival. GEO appears in 67 of 72 answers (93%), AEO in 76%, “AI visibility” in 43% (72% of ChatGPT’s answers; it is the category name of tracking vendors such as Profound and Peec AI), AI SEO in 28%, AI search optimization in 22%, LLMO in 19% and AIO in 10%.
• Fan-outs are invisible to keyword tools. Of 89 distinct fan-out queries, 46 exceed the keyword database’s limits (10 words, 80 characters). Of the other 43, three have Google volume: “what is generative engine optimization” (1,300 at $12.70), “best answer engine optimization tools” (210) and “difference between seo aeo and geo” (20 at $9.25). ChatGPT’s searches average 12.8 words, and 29 of its 56 add a year.
• Google’s own documentation is the most-cited source. developers.google.com was cited in 20 of 72 answers, blog.hubspot.com in 13, Search Engine Land in 12. Wikipedia was cited in 5 (Gemini 2, Claude 3, ChatGPT 0). In a single probe outside the panel, Gemini cited the Wikipedia GEO article first and opened by calling GEO “also sometimes called AI Optimization (AIO) or Answer Engine Optimization (AEO)”: the aliases of the two articles Wikipedia merged into GEO on February 26, 2026.
• Small panel. One to three runs a prompt; fan-outs vary from run to run (ChatGPT’s 3 runs of the same prompt never produced identical query sets). Read the shares as directions, not rates.