Last updated: 15 October 2026
GEO (generative engine optimisation — the practice of shaping content so AI tools cite it) does work in controlled testing: a KDD 2026 benchmark of 10,000 queries found certain techniques lifted visibility by up to 40%. But no study has yet proven these gains reliably convert into clicks, leads or revenue for a UK business over time.
Key Takeaways
- The KDD 2026 study by researchers from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi found GEO techniques can boost visibility in generative engine responses by up to 40% across a 10,000-query benchmark (Aggarwal et al., KDD 2026).
- Quotation Addition was the single best-performing tactic in that study, lifting Position-Adjusted Word Count by 41% and Subjective Impression by 28% versus baseline (Aggarwal et al., KDD 2026).
- Keyword stuffing actively hurts visibility, scoring 8% below baseline in the benchmark and 10% below baseline in live Perplexity testing (Aggarwal et al., KDD 2026).
- A 2026 critical survey found no GEO technique with a stable, cross-platform causal effect on organic discoverability, clicks or conversions over time (Critical Survey of GEO, 2023-2026).
- Aether AI published 281 articles across four client brands in a single 30-day period in September 2026, using citation tracking across six AI engines to measure what actually gets quoted rather than relying on theory alone.
What is GEO and how does it differ from traditional SEO?
GEO (generative engine optimisation) is the practice of structuring and writing content so that generative AI systems — ChatGPT, Google AI Overviews, Perplexity, Claude, Gemini and Microsoft Copilot — select, cite and quote it when answering a user's question. Traditional SEO, by contrast, optimises for ranking a page in a list of blue links on a search engine results page.
The distinction matters because the end product is different. SEO succeeds when a page ranks first on Google; GEO succeeds when a sentence from that page is extracted into an AI-generated answer, often with no click required at all.
James Scherer, VP of Strategy at Foundation Inc, frames it this way: "SEO is your space — your website, blog, technical optimization. GEO is all that stuff plus external influences. We don't really optimize for generative engines; we influence them." That's a meaningful admission from inside the industry — GEO is closer to earning a citation than controlling a ranking.
Google itself has pushed back on the idea that GEO requires entirely new machinery. Its own documentation states plainly: "You don't need to create new machine readable files, AI text files, markup, or Markdown" (Google's Guide to Optimizing for Generative AI Features on Google Search). In other words, GEO builds on sound SEO fundamentals — crawlability, structure, clarity — rather than replacing them.
Is there credible evidence that GEO improves visibility in AI answers?
Yes, but the evidence is narrower than most marketing claims suggest. The foundational study — Aggarwal et al.'s "GEO: Generative Engine Optimization," presented at KDD 2026 by researchers from Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi — built a 10,000-query benchmark called GEO-bench and tested specific content changes against it.
The headline finding: certain optimisation methods, such as adding citations, quotations and statistics, can boost visibility in generative engine responses by up to 40% (Aggarwal et al., KDD 2026). That's a genuinely significant number from a peer-reviewed benchmark, not a vendor claim.
But a 2026 academic survey complicates the picture considerably. Reviewing GEO literature published between 2023 and 2026, it found no technique with a stable, longitudinal, cross-platform causal effect on organic discoverability or downstream clicks and conversions — even though content that has already been retrieved by an AI system can demonstrably influence the final answer (Optimizing Visibility in Generative Engines: A Critical Survey, 2023-2026).
Put simply: GEO can change which sentence an AI model picks once your page is already in its retrieval set. Whether any technique reliably gets your page into that set in the first place, across ChatGPT, Perplexity, Google AI Overviews and Claude simultaneously, remains unproven. Independent practitioner critiques of the original KDD methodology — including a detailed breakdown on Sandbox SEO — raise further questions about how far the lab results generalise to live, commercial websites.
There's also the attribution problem. Pew Research found users clicked through on just 8% of searches when an AI Overview was present, versus 15% without one (Pew Research, cited by Similarweb). Even a successful citation may never show up as a click in Google Analytics — it often arrives as "Direct" traffic, or no traffic at all.
Which GEO techniques are backed by testing versus speculation?
Quotation Addition and Statistics Addition are the two techniques with the strongest experimental backing; persuasive tone and keyword stuffing are not. The KDD 2026 benchmark tested specific, named interventions rather than vague "best practices," and the gap between the best and worst performers was substantial.
Quotation Addition — inserting relevant, sourced quotations into content — was the top performer, improving Position-Adjusted Word Count by 41% and Subjective Impression by 28% over baseline. Statistics Addition — citing concrete figures with sources — came second, improving the same metrics by 31% and 23% respectively (Aggarwal et al., KDD 2026).
At the other end, keyword stuffing scored 8% below the unmodified baseline on Position-Adjusted Word Count in the benchmark, and a full 10% below baseline when validated live against Perplexity.ai (Aggarwal et al., KDD 2026). That's not a neutral result — it's actively counterproductive, echoing two decades of SEO guidance that over-optimised copy reads as spam to both humans and machines.
Equally telling: testing a more persuasive or authoritative tone showed no significant improvement in visibility, suggesting generative engines are already fairly robust to rhetorical flourishes (Aggarwal et al., GEO: Generative Engine Optimization). You cannot talk your way into a citation; you need substantiated, well-sourced content.
| Technique | Tested effect | Verdict |
|---|---|---|
| Quotation Addition | +41% Position-Adjusted Word Count, +28% Subjective Impression | Strongest evidence |
| Statistics Addition | +31% Position-Adjusted Word Count, +23% Subjective Impression | Strong evidence |
| Citing sources (lower-ranked sites) | +115.1% visibility for 5th-ranked sites | Strong, rank-dependent |
| Citing sources (top-ranked sites) | -30.3% visibility for 1st-ranked sites | Can backfire |
| Persuasive/authoritative tone | No significant improvement | Speculative |
| Keyword stuffing | -8% to -10% vs baseline | Counterproductive |
One of the more surprising findings concerns site authority. Fifth-ranked websites gained +115.1% visibility from adding source citations, while first-ranked sites actually lost 30.3% visibility from the identical tactic (Peec AI, citing Aggarwal et al. KDD 2026). GEO appears to help smaller or lower-authority sites close the gap far more than it helps established market leaders — useful news for SMEs across the UK who assume GEO is only a game for large incumbents.
How long does it take to see measurable results from GEO?
There is no fixed, research-backed timeline for GEO results, because no longitudinal cross-platform study has established one. The 2026 critical survey's central finding — that no technique shows a stable causal effect over time across platforms — is itself the honest answer to "how long will this take" (Critical Survey of GEO, 2023-2026).
In practice, most UK teams running GEO alongside SEO should expect a similar rhythm to content marketing generally: indexing and crawling happen within days to weeks, but citation frequency in AI answers tends to shift gradually as models refresh their retrieval indexes and as more of your content accumulates the citation-friendly signals (quotations, sourced statistics, clear structure) that the KDD research found effective.
Aether AI's own operational data offers a useful data point here. Across four client brands — spanning security, facilities software and branding — Aether AI published 281 articles in the 30 days to September 2026, tracking citations across six AI engines simultaneously (ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Microsoft Copilot) rather than waiting months for a single annual review. That cadence is consistent with the KDD finding that visibility responds to content-level changes, and it means issues — a technique that isn't landing on one engine, say — surface in weeks rather than quarters.
"Vendors will argue AEO versus GEO versus LLM SEO forever because category names are marketing territory. Buyers shouldn't care. The question underneath is identical: when an AI assistant answers a question in your market, are you in the answer, and can you prove it?" — Lauren Dawkins, Head of Content, Aether AI
That question — can you prove it — is precisely why citation tracking matters more than theoretical timelines.
Who should own GEO responsibilities within a business?
GEO responsibility should sit with whoever already owns content and SEO within a business, not a separate new hire, because the underlying skills — research, sourcing, structure, crawlability — overlap almost entirely with existing SEO and content disciplines. Digiday's reporting on the category found experts broadly describing GEO as "more SEO than new discipline" (Digiday, GEO hype busted), which argues against building a standalone team from scratch.
That said, GEO does introduce one genuinely new operational need: citation tracking across multiple AI engines, which traditional SEO tools (built around Google rankings) were never designed to do. For most UK marketing teams, that means either upskilling an existing SEO or content lead, or adopting a platform purpose-built for multi-engine citation tracking rather than hiring a dedicated "GEO specialist" with no established qualification pathway or professional body behind the title.
What are the common mistakes businesses make attempting GEO?
The single most common mistake is treating GEO as a rebrand of keyword-stuffed SEO copy rather than as a citation-earning discipline, which the KDD 2026 data shows actively backfires. Keyword density tactics that once nudged Google rankings now score 8-10% below an unmodified baseline in AI visibility testing (Aggarwal et al., KDD 2026).
A second mistake is chasing novelty over fundamentals — adding llms.txt files, exotic markup or AI-specific sitemaps on the assumption that these are the lever that matters.
"Treat llms.txt like you treated meta keywords: cheap to add, unproven as a lever, no substitute for the fundamentals. What measurably matters is whether AI crawlers can fetch and parse your pages at all — server-rendered content, clean structure, no login walls in front of your best answers." — Lauren Dawkins, Head of Content, Aether AI
Other recurring errors include:
- Measuring only rankings, not citations — rank position in Google tells you nothing about whether ChatGPT or Perplexity quoted your page.
- Ignoring brand mentions off-site, despite research showing 86% of citations in AI-generated answers come from brand-controlled or brand-adjacent sources, and a separate study finding a 65% correlation between brand mention frequency in web content and in AI-generated responses (Yext research / Seer Interactive study, cited by Similarweb).
- Writing for persuasion rather than substantiation — the research found tone changes delivered no measurable lift, while quotes and statistics did.
- Assuming GEO replaces SEO rather than complementing it — The Drum's analysis of GEO myth-busting makes the same point about misplaced "GEO versus SEO" framing (The Drum).
How is GEO success measured, and what should a business check first?
GEO success is measured primarily through citation tracking — monitoring how often and in what form a brand's content is quoted across AI engines — rather than through traditional keyword rank position. Google Search Console still matters for the crawlability and indexing layer GEO depends on, but it cannot tell you whether ChatGPT cited your page.
Before investing, a UK business should check whether its actual customers are using AI assistants to research the category at all. A B2B professional services firm serving procurement teams in Leeds or Manchester may see very different AI-search behaviour from a consumer e-commerce brand. Checking Google Search Console for existing organic queries, reviewing competitor visibility across engines, and running a free AI-visibility audit are sensible first steps before committing budget.
Comparison: measuring GEO in-house vs using a dedicated platform
| Approach | What it covers | Limitation |
|---|---|---|
| Manually querying ChatGPT/Perplexity by hand | Spot-checks specific queries | Not scalable, no historical tracking, single-engine at a time |
| Google Search Console alone | Crawling, indexing, organic clicks | Cannot see AI engine citations at all |
| Multi-engine citation tracking platform (e.g. Aether AI) | Citations across six AI engines, keyword and competitor tracking, GSC integration | Requires platform adoption and ongoing monitoring |
Aether AI's own platform tracks citations across all six major engines — ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot — precisely because the research above shows visibility behaves differently on each one, and a single-engine spot-check will miss most of the picture.
Your GEO checklist
- Audit whether your target customers actually use AI assistants to research your category before committing budget.
- Add sourced statistics and direct quotations to existing high-performing pages rather than rewriting from scratch.
- Remove keyword-stuffed copy — the KDD 2026 data shows it scores below an unoptimised baseline.
- Check server-rendered content is accessible to AI crawlers, with no login walls in front of key answer pages.
- Track citations across at least three AI engines, not just Google rankings.
- Monitor brand mentions on third-party sites, not just your own domain.
- Treat llms.txt and similar files as low-cost extras, never as a substitute for crawlable, well-structured content.
- Review citation data monthly rather than waiting for an annual SEO audit.
FAQ
Does GEO actually work for improving AI search visibility?
Yes, in controlled testing: the KDD 2026 benchmark found specific techniques improved visibility by up to 40% (Aggarwal et al., KDD 2026). A 2026 critical survey, however, found no technique with a proven stable effect on real-world clicks or conversions over time, so results vary by platform and context.
What is the difference between GEO and SEO?
SEO optimises content to rank in search engine results pages; GEO optimises content to be cited or quoted inside AI-generated answers. The two overlap heavily on fundamentals like crawlability and clear structure, but GEO adds citation tracking across AI engines as a distinct new measurement need.
Can GEO tactics backfire or hurt visibility?
Yes. Keyword stuffing scored 8-10% below baseline in both benchmark and live Perplexity testing (Aggarwal et al., KDD 2026), and source citations actually reduced visibility by 30.3% for already-top-ranked sites (Peec AI).
Does keyword stuffing help or hurt generative engine visibility?
Keyword stuffing hurts visibility. It scored 8% below the unmodified baseline in the KDD 2026 benchmark and 10% below baseline in real-world Perplexity.ai validation testing.
How do you measure whether GEO is working for your brand?
Track citation frequency across multiple AI engines — ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Copilot — rather than relying on Google rank position alone, since 86% of AI citations come from brand-controlled or brand-adjacent sources (Yext research, cited by Similarweb).
Is GEO just a rebranded version of SEO?
Largely, yes, according to industry commentary: Digiday reported experts describing GEO as "more SEO than new discipline" (Digiday). GEO builds on SEO fundamentals and adds AI-specific citation measurement rather than replacing established practice.
Do AI Overviews and zero-click search reduce the value of GEO?
They change how value is captured rather than eliminating it. Pew Research found click-through dropped to 8% of searches with an AI Overview present versus 15% without one (Pew Research, cited by Similarweb), meaning citation and brand visibility matter even when a click never happens.
Is GEO worth investing in for small businesses or lower-ranked websites?
The evidence suggests yes, more than for established leaders. Fifth-ranked sites gained 115.1% visibility from source citations while first-ranked sites lost 30.3% from the same tactic, showing GEO can narrow the gap for smaller players (Peec AI, citing Aggarwal et al.).
Measuring your own GEO performance with Aether AI
Every question in this article — which techniques work, how long results take, how to measure success — comes down to one practical problem: most businesses have no visibility into whether AI engines cite them at all. Aether AI was built specifically to close that gap, tracking citations across ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini and Microsoft Copilot in one place, alongside keyword and competitor tracking and Google Search Console integration.
Aether AI is the same engine the agency runs for its own clients — Priority First, Aether Agency and Pulse Operations among them — publishing 281 articles across four brands in the 30 days to September 2026 and measuring every one against the six-engine citation data described above, rather than relying on theory alone.
If you're uncertain whether GEO is relevant to your industry or customer search behaviour, start with the free AI-visibility audit at /audit to see where your brand currently stands before committing further budget.