SEO and GEO: the types that matter, and what actually still works

Ask three SEO tools how many types of SEO exist and you get three different answers. Semrush names three core types and five more. Ahrefs names three core types and sixty-five more. GeeksforGeeks lists twelve with no hierarchy between them at all. None of them agree, and Google's own documentation does not use this taxonomy at all: it talks about technical requirements and content quality, not "on-page" or "off-page" anything. That disagreement matters more now than it used to, because a second question has been bolted onto the first: what is GEO, and does it replace any of this?

The types that most sources agree on

Two sources checked here, Semrush and Ahrefs, independently name the same three types as core: on-page SEO, the content and structure changes made to a page itself; off-page SEO, the external signals, chiefly backlinks, that build a site's authority; and technical SEO, the crawlability, indexing and site-structure work that lets a search engine find and understand a page in the first place. Local SEO, aimed at businesses that want to be found by people nearby, sits outside that core three for both of them, in its own category.

GeeksforGeeks disagrees with the hierarchy, not the definitions: its list of twelve types puts local, ecommerce, video, image, international, voice, YouTube, social and content SEO alongside on-page, off-page and technical, with no core-versus-secondary distinction at all. The practical difference between "eight secondary types" and "twelve equal types" is mostly how a guide chooses to organise the same underlying work, not a disagreement about what the work is.

For a business without an in-house SEO team, the version worth acting on is closer to the following, since these are the categories that actually change what gets built or written:

• On-page SEO: what is on the page itself, titles, headings, the words used, and whether they answer what someone actually searched for.
• Off-page SEO: what other sites say about this one, chiefly links and mentions from sources worth being associated with.
• Technical SEO: whether a search engine can reach, read and index the page at all, before any of the above matters.
• Local SEO: whether a business shows up for people searching nearby, which depends on accurate business listings as much as the website itself.
• Content SEO: whether the page is worth ranking, judged on whether it answers the question completely and honestly rather than padding it out.
• Image and video SEO: whether visual content is described well enough, through file names, alt text and captions, for it to be found and understood on its own.

Google's own Search Essentials documentation, last updated December 2025, never uses any of these terms. It describes "technical requirements" (a crawlable site returning a normal response, indexable content) and "key best practices" (helpful, people-first content), and leaves the vendor taxonomy out entirely. That is not a contradiction. It is Google confirming that the requirements underneath the labels are what earns a ranking, and the labels are a way the industry organises advice about meeting them, not a scoring system Google itself uses.

What GEO actually is

Generative engine optimization is a specific, traceable term. It comes from a paper published to arXiv in November 2023 by researchers at IIT Delhi, Princeton and independent researchers, presented at the KDD 2024 data-mining conference. The paper defines GEO as optimizing content so generative systems select and cite it when composing an answer from multiple sources, rather than only ranking it as a link, and reports that specific tactics, adding citations, statistics and quotations among them, improved a source's visibility in its own benchmark by up to 40%. That figure is nearly two years old and was measured before the current generation of AI Overviews and ChatGPT search existed, so treat it as evidence that the underlying idea works, not as a number this year's content will reproduce.

What GEO is not is a settled discipline with its own rulebook. Wikipedia's summary of the term states plainly that no consensus definition exists separating GEO from the older term AEO, answer engine optimization, or from AI optimization and LLM optimization, terms some vendors use for the same thing. HubSpot's own comparison of AEO and GEO reaches the same conclusion: the practical difference is mostly terminological, and most marketers use the two interchangeably. Google's stated position, as reported on the same Wikipedia page, is more direct still: optimizing for generative AI search is optimizing for the search experience, and is therefore still SEO. Anyone selling a hard line between the two is selling a distinction the platforms themselves have not drawn.

What Google actually says to do, and not do

This is the part most guides skip, because Google's own documentation is unusually direct about it, and it does not support the tactics that get repeated most often.

llms.txt does not help. Google's AI-optimization guide states it in as many words: Google Search ignores llms.txt files, and adopting one will neither help nor harm a site's visibility or rankings. This has not changed; it was true in July 2026, when the guide was last updated, and it is still true now.

No special AI schema exists. The same guide says structured data is not required for generative AI features and there is no special schema.org markup to add. Structured data still helps a page qualify for rich results in ordinary search, which is worth doing for its own sake, but it is not a citation hack.

Content does not need rewriting "for AI." Chunking content differently, pursuing artificial mentions across the web, and obsessing over long-tail keyword variants are all named directly in Google's guidance as unnecessary, and creating near-duplicate pages to chase query variants can trigger its scaled-content spam policy rather than help.

The mechanism runs through ordinary search first. Google states this as plainly as anything else in the guide: the way Search finds and processes pages is the core of how its AI systems access data at all. A page has to be crawlable, indexable and eligible to appear in a normal result before any generative feature can draw on it. There is no separate AI front door.

What is actually rewarded is what Google has said for years: original, well-sourced, honestly authored content. Google's helpful-content documentation asks whether a page offers original information, research or analysis rather than summarising what others have already said, and it "strongly encourages" visible authorship, a byline a reader can check. None of this is new advice dressed up for AI. It is the same advice, now with a more literal reason to follow it: a system that is trying to decide what to cite is applying the same test a careful reader would.

What actually changed, and what did not

The scale is real, and worth being specific about rather than repeating a rounded headline figure. Google's own investor materials put AI Overviews at over 2.5 billion monthly users and AI Mode past 1 billion, with the Gemini app at 950 million monthly active users, as of its mid-2026 reporting. OpenAI has confirmed 900 million weekly active users for ChatGPT and 50 million paying subscribers, disclosed alongside its funding announcement in February 2026. Pew Research's own February 2026 US survey found 44% of American adults had used ChatGPT, up from 34% the year before and 18% in 2023, and that 49% now use an AI chatbot of some kind. In the UK specifically, Ofcom's News Report 2026 found 21% of UK adults had used an AI app or service for news in the past month, with ChatGPT the single most-used tool for that at 13%. These are the actual, sourced numbers; treat any AI-search statistic that arrives without a named report and a date as unverified, because most of what circulates does.

What has not changed is which reporting exists to measure any of this from a website's own side. Google's generative AI performance report, fully rolled out worldwide at the end of August 2026, shows impressions only, broken down by page, country, device and date, with no clicks or query data. Bing's AI Performance report is citations-based and gained analytical add-ons in June 2026, but is still built around the same idea: whether a page was shown as a source, not whether it drove a click. Reporting on this is genuinely behind the scale of the shift, which is exactly why a claimed precise measurement of "AI visibility" from a third party is worth reading carefully rather than taking at face value.

Where this leaves a business trying to be found

Nothing here replaces the fundamentals: a crawlable, correctly indexed site, content that actually answers what someone is asking, and enough external signal that a search engine, or a generative system doing the same underlying work, has reason to trust it. What changes is the audience for that content now includes a system composing an answer, not only a person scanning a results page, and the evidence for what that system rewards points at the same honesty and depth Google has been asking for since long before AI Overviews existed. Our GEO vs SEO guide sets out exactly what that comparison does and does not change, and our LLM SEO guide covers the crawler-access mechanics behind being cited by ChatGPT, Claude and Perplexity specifically.

Checking what currently ranks for these questions is itself instructive. Search results for "GEO agency UK" are entirely commercial: agency service pages and other agencies' "best GEO agency" listicles, with no independent journalism, consumer body or government source anywhere in the top results. Search results for the plain definitional question, "what is GEO", are more mixed: an academic paper, a vendor blog, Wikipedia, another paper, an agency blog, and a venture-capital firm's own blog post, none of them citing Google's or Bing's own documentation directly. That gap, primary guidance being the thing missing from a market full of opinions about it, is the reason this piece leads with what Google's own guide says rather than repeating what other blogs say about it.

That is the work behind GEO and SEO: finding the questions a business's actual buyers are asking, checking whether the site currently answers them well enough to be cited, and fixing the technical and content gaps that stop it. If your own search visibility has not been looked at since before AI Overviews existed, talk to us about what an honest assessment would actually check.

Written by

Michael Szalaty Technical director

Delivery, development and infrastructure. Runs a staffing agency.