Two of these three acronyms name the same job. No dated document stands behind the third. Sorting them out takes about twenty minutes with the primary documents open...the order that falls out is not the one most vendor pages hand over.

Which Is Which

GEO and AEO describe the same work from two angles. Generative engine optimization is named for the system producing the answer...answer engine optimization is named for what the reader ends up holding.

  • GEO: Traceable to a paper accepted at KDD 2024 with six named authors.
  • AEO: Carried into a shipped product name by HubSpot in April 2026.
  • AIO: Absent from Google's guidance and from any document carrying a date.

Where GEO Came From

GEO is the one acronym that traces to a document with a date and a byline. "GEO: Generative Engine Optimization" reached arXiv on 16 November 2023, was revised twice through 2024, and was accepted to KDD 2024 under National Science Foundation Grant No. 2107048. The byline runs six names, Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan and Ameet Deshpande.

Three affiliations are printed on it, the Indian Institute of Technology Delhi, Princeton for three authors, and Independent, Seattle for two. One vendor explainer calls it a Princeton and Georgia Tech paper. Georgia appears nowhere in its full text. Two of the authors have held Georgia Tech affiliations on other work...that reads as carryover rather than invention.

Wikipedia's article on the term does not cite the paper at all...its earliest attribution for the concept is John Herrman's Intelligencer piece from 4 August 2025, nearly two years after the arXiv submission.

Nothing comparable exists for AIO. No dated, authored document could be found putting those three letters to visibility work...vendor-coined is the honest description. Nothing dated surfaced for AEO either, and an absent record is not proof of none.

What The 40% Measured

That paper's abstract claims its methods "can boost visibility by up to 40% in generative engine responses"...that one sentence carries more commercial weight than the rest of the category combined. Four features of the experiment decide how much it can actually hold.

  • Engine: A simulator built on GPT-3.5-turbo, not a live commercial assistant.
  • Sources: The top five Google results for each query, fetched in advance.
  • Benchmark: GEO-bench, 10,000 queries split 8,000 train, 1,000 validation, 1,000 test.
  • Metric: Position-Adjusted Word Count, weighted by where in the answer the wording lands.

Look hard at the last row, because it is the one that gets mangled. Position-Adjusted Word Count measures how much of a finished answer's wording is attributed to a given source, which is a share of the output. It is not a probability of being cited...it could not be, because every document in the test was already sitting in the model's context before measurement began.

Its result sentence reports "a relative improvement of 30-40% on the Position-Adjusted Word Count metric and 15-30% on the Subjective Impression metric". That range belongs to three methods together, Cite Sources, Quotation Addition and Statistics Addition. Rounding it to a bare 40% takes the top of one range, then staples it to whichever tactic is being sold.

Just one live engine was tested. Running the same methods against "Perplexity.ai, a real-world generative engine" produced "visibility improvements up to 37%", close enough to the headline to matter. Limitations get a blunter treatment in the paper than in any vendor summary of it...owing to the black-box nature of search engine algorithms, the authors state, they did not evaluate how GEO methods affect search rankings.

Inside The Results Table

A critical survey of 45 GEO studies, a single-author preprint by Olivier Martinez submitted 15 July 2026, reproduced the original table and traced the famous number to one cell.

  • Baseline: 19.3, the reference condition every other row moves against.
  • Quotation Addition: 27.2, roughly 41% above baseline and the highest row in the table.
  • Statistics Addition: 25.2, roughly 31% above baseline on the same metric.
  • Keyword Stuffing: 17.7, below the baseline it was supposed to beat.

Martinez states verbatim that the result "does not mean that 40% more readers will click, nor that a page will gain 40% in retrieval probability". Martinez also names the structural catch. All five sources' shares "sum to one", so the measure is "intrinsically relative and redistributive"...one source gains exactly what the others give up.

Naming The Misquote Plainly

So 41% belongs to Quotation Addition and 31% to Statistics Addition...the two are not interchangeable. Vendor pages that recommend adding statistics and then cite content as "up to 40% more likely to be cited" have made two errors in one clause. They moved the larger number onto the smaller tactic...they converted a share of an answer's wording into a probability of appearing at all.

Anyone can check that against the table above, and nothing a buyer learns about this category is worth more.

Which Letters Ship On Products

Vendor positioning splits the three labels unevenly...it does not split along the evidence. Note where the paper sits, because it is not behind the label most platforms put on the box.

  • Conductor: Sells itself as "#1 for Enterprise AEO" and the only unified platform for it.
  • Optimizely: Announced a "full AEO platform" on 10 June 2026, built with Conductor.
  • Adobe: Announced on 14 October 2025 a solution labelled "Generative Engine Optimization (GEO)".
  • Semrush: Close to alone among the large platforms shipping product under the AIO name.

Profound picked AEO on the record in a post dated 29 June 2025, and its stated reasoning is about the word rather than about the work. "Distinctiveness: AEO is unique in search marketing", the post says, before noting that "GEO conflicts with geography, geology, and geo-targeting" and closing with "good luck owning that acronym."

Crowding in a three letter string is fair for a business to weigh. It is also not evidence about how answers get assembled...a buyer reading vendor labels as a taxonomy is reading trademark strategy instead.

Recognition Versus Actual Use

Usage was measured twice at Search Engine Land, both times by Kelsey Libert of Fractl, and GEO leads on every reading. The first, published 21 November 2025, set 342 reader survey respondents alongside 33,250 job postings on Indeed, 6,400 LinkedIn posts and 3,350 Reddit threads from August to October 2025.

  • Recognition: GEO at 84% among marketer respondents, against 61% for AEO.
  • Forced Choice: GEO at 42% as the single term used, against 14% for AEO.
  • Trend Movement: GEO up 121% quarter over quarter on Google Trends, against 56% for AEO.

The second, published 22 January 2026 with Danny Goodwin, read every 2025 LinkedIn post from 75 leading SEO practitioners and scored each on a scale running minus one to plus one. Across that set GEO turned up in 59% of the posts, with 82% of those scoring positive.

One figure in that study cuts against the rest. Three percent of those practitioners' LinkedIn headlines reference GEO, against the 84% who recognised the term when a survey asked...what people know is not what they will hang next to their own name.

AEO Went Into Product

HubSpot moved AEO from vocabulary into a product name on 14 April 2026, the strongest commitment any of the three labels has received.

  • Reason: CEO Yamini Rangan, quoted in the announcement, says "how buyers search is fundamentally changing".
  • Number: The same release states organic traffic for HubSpot customers is down 27% year over year.
  • Scope: Its blog, updated 22 May 2026, uses AEO "to describe all motions used to improve citations and mentions in answer engines".
  • Concession: The same post grants that "some industry experts use GEO to refer to how AI generates responses based on prompts, and use AEO as a broader umbrella term".

That concession is the useful part. A company that has just staked a product name on AEO states in its own documentation that the AEO and GEO hierarchy is contested...that is roughly as settled as this dispute currently gets.

Google Names None Of Them

Google runs AI Overviews and AI Mode...its guidance for site owners uses none of the three. Its page for them, titled "AI features and your website" and last updated 10 December 2025, contains the strings GEO, AEO and AIO nowhere at all. It rules the premise out as well, stating there are "no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary".

Google's page skips this industry's main noun too. Nowhere on it does the word citation appear, and what it describes showing is "a wider and more diverse set of helpful links"...the vocabulary every dashboard in the category runs on was supplied by the people selling the dashboards.

One other Google document does name two of the three. Its AI optimization guide spells out that "'AEO' stands for 'answer engine optimization' and 'GEO' for 'generative engine optimization'", then answers that optimizing for generative AI search is "optimizing for the search experience", and still SEO. Further down, the same guide tells readers to "Prioritize effective SEO strategies over 'AEO/GEO hacks'". AIO appears nowhere near that passage.

Read the two pages together and Google has ranked the labels without intending to. Two were worth naming in order to disagree with...the third never came up.

OpenAI publishes nothing comparable. Documentation for its crawlers sets out which bots fetch what and how to keep them out, and says nothing about how a fetched page gets selected for an answer. So half the mechanism claims in this category are inference from watching citations, not readings of a published system.

A Startup Sized Hole

Acronyms multiply here for a reason...it is not a technical one. Forrester's Nikhil Lai wrote on 24 November 2025 that "AEO is significantly, but not fundamentally, different from SEO", and that point solutions popularising new acronyms "tend to exaggerate SEO and AEO's differences to carve a startup-sized hole in marketers' tech stacks".

His post names six such vendors, Profound among them, and carries no data at all, no study, no sample and no measurement. Forrester sells research subscriptions...treat the piece as a well placed opinion rather than as a finding.

A looser problem sits under all three labels rather than between them. Martinez counts the word visibility standing in for at least nine different quantities, activation, retrieval, mention, citation, prominence, coverage, absorption, fidelity and behavior, each of which "addresses a distinct question and has its own denominator".

Three letters cannot tighten a noun that loose. Choosing between GEO, AEO and AIO settles what to call the invoice...it settles nothing about what the number printed on it counted.

What To Call It

A layer diagram will be obsolete by the next product launch...a rule will not.

  • Write GEO by default. Only that acronym carries a dated, authored coinage behind it.
  • Say AEO when the buyer says AEO. HubSpot, Conductor and Optimizely all ship products carrying those letters.
  • Never say AIO without expanding it. That argument is laid out on its own page.

Whatever the label settles into, the check underneath does not change. vizmeta HQ at hq.vizmeta.com asks ChatGPT, Claude and Google AI once a month whether they name a given business, and logs the answers so a trend sits on a page.

Arguments about the labels will keep running...a buyer can sit them out. Ask which document a claim rests on, then open it and read the sentence being quoted.