Google says optimising for its AI features is still SEO, and its own documentation lists what you can skip. But Google is not the only engine. Across the same 8 Indian agency sites we measured, Google AI Mode returned 971 citations and ChatGPT 27. GEO is a distribution question.
Our view is that this question usually gets answered by someone with something to sell, in one of two directions. Either GEO is a new discipline you must buy today, or it is a rebrand of work you already pay for. The documentation and the research support a narrower answer than either.
Does Google treat GEO as different from SEO?
No, and it says so in the guide it published for exactly this question.
"AEO stands for answer engine optimization and GEO for generative engine optimization. These are both terms you may see used to describe work specifically focused on improving visibility in AI search experiences. From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."
That is Google's guide to optimising for generative AI features, last updated 10 July 2026. The reasoning is structural, not rhetorical. Google's AI features retrieve from the ordinary Search index, so the thing being optimised is the same index either way.
Three conditions gate eligibility, and the third is easy to miss:
- 1
The page is indexed
- 2
The page is eligible to show with a snippet
- 3
The site is included in Search generative AI features in Search Console
What did the paper that coined the term actually find?
Generative engine optimization is not a vendor coinage. It comes from a paper by Aggarwal and colleagues at IIT Delhi and Princeton, presented at KDD 2024, which tested nine content changes against a benchmark of 10,000 queries and then against Perplexity.ai.
| Method | Position-adjusted word count | Verdict |
|---|---|---|
| Quotation Addition | 27.2 | Best overall |
| Statistics Addition | 25.2 | Strong |
| Cite Sources | 24.6 | Strong |
| No optimisation | 19.3 | Baseline |
| Keyword Stuffing | 17.7 | Worse than doing nothing |
On Perplexity.ai the pattern held harder still. Keyword stuffing, the classic SEO move, performs 10% worse than the baseline.
The widely repeated 40% lift needs care. It is the authors' best case on a benchmark of 10,000 queries run against their own simulated engine, measured as share of words inside a generated answer, not traffic and not rankings.
"Owing to the black-box nature of search engine algorithms, we didn't evaluate how GEO methods affect search rankings."
Which engine actually cites you?
Google's position holds for Google's surfaces. It says nothing about the others, so we measured them. We took every domain ranking in the Google India top 10 for five marketing-services queries, 28 domains after deduplication, and pulled citation counts across seven AI engines.
| Engine | Citations, 8 Indian agency sites |
|---|---|
| Google AI Mode | 971 |
| Google AI Overviews | 352 |
| Perplexity | 193 |
| Gemini | 135 |
| Grok | 52 |
| Copilot | 46 |
| ChatGPT | 27 |
Same eight sites, same day, a 36 to 1 gap between the top engine and ChatGPT. 3 of 8 have no ChatGPT citation at all, which is 37.5%.
See how you show up in ChatGPT, Gemini and Perplexity. Free AI visibility audit.
Does company size change which engines find you?
Sharply, in both our data and the research.
- Full set: 23 of 28 domains cited by all seven engines, which is 82.1%
- Smaller agency sites: only 4 of 8 cited by all seven, which is 50.0%
The KDD authors measured a related but distinct asymmetry, by search position rather than company size, and found it inverted:
| Site's Google position | Effect of adding citations |
|---|---|
| Ranked fifth | Gained 115.1% visibility |
| Ranked first | Lost 30.3% visibility |
That is the most useful finding in the paper for anyone not already winning. These tactics move the challenger and can cost the incumbent, which inverts the usual assumption that the market leader gains most.
What does Google say you can skip?
The same guide lists tactics it says do nothing for Google Search. This matters because most GEO packages sell at least one of them.
- LLMS.txt files and other AI-specific markup, which Google Search ignores
- Chunking content into small pieces for machines
- Rewriting existing content purely for AI systems
- Seeking inauthentic mentions across the web
- Adding structured data specifically to reach AI answers, though it stays worth having for rich results
Engine behaviour does genuinely differ, and that is the honest core of the GEO case rather than any of the above. Perplexity publishes its own separate crawler rules, and notes that its user-triggered fetcher generally ignores robots.txt rules. Our breakdown of the crawler documentation compares all of them side by side.
So how should you decide?
Treat this as one question about distribution, not two budgets. The order matters more than the effort.
- 1
Check whether your buyers' queries even return an AI answer
- 2
Check whether the engines they use cite you today
- 3
Fix crawlability, because nothing else matters if the page cannot be read
- 4
Only then change content, preferring quotations, statistics and citations
Steps one and two are free and take an afternoon. Our post on why a site does not show in AI search walks those checks.
Our AI Overviews trigger data covers which query types produce an AI answer at all.
If you would rather someone else ran it, that is what our AI visibility audit is.
What this cannot tell you
Our counts come from Ahrefs' AI citation index, not from us prompting the engines ourselves. A zero may mean an engine does not cite that domain, or that coverage for it is thin. We have not ruled the second out.
Gemini is the clearest case, and the reason we left it out of the headline:
| Gemini check, 16 largest domains | Value |
|---|---|
| Median Gemini count as share of ChatGPT | 25.8% |
| One reference site, ChatGPT citations | 43,889 |
| Same site, Gemini citations | 1 |
This is one snapshot, not a trend. The entry criterion was ranking in Google India for a marketing-services query, so it says nothing about other industries.
Key takeaways
- For Google's own AI features, GEO is SEO. Google states this directly
- The engines disagree enormously. 971 citations against 27, on the same 8 sites
- Keyword stuffing measured worse than no optimisation on a real generative engine
- These tactics favour the challenger. Rank five gained 115.1%, rank one lost 30.3%, n=10,000 queries
- Check crawlability and demand before buying any GEO package
Frequently asked
Is GEO replacing SEO?
No. Google's documentation says optimising for its generative AI features is optimising for the search experience, and thus still SEO, because those features retrieve from the same Search index. What is genuinely new is that other engines exist and behave differently, which is a distribution question rather than a new discipline.
Do I need a separate GEO budget?
Probably not a separate one. The work that helps is crawlability, unique content, and the quotation and statistics patterns the KDD research found effective, all of which are ordinary content and technical work. A separate budget makes sense only if a non-Google engine is where your buyers actually are.
Does llms.txt help?
Not for Google Search. Google states it does not use these files, and that creating them will neither harm nor help visibility there. Other services may use them, so it is not harmful, just not a lever for Google.
Which AI engine should we care about in India?
On our sample, Google AI Mode cited these sites far more than any other engine, 971 times against 27 for ChatGPT across 8 domains. That covers marketing-services queries in India only. Measure your own category before assuming it transfers.
How do we measure AI visibility ourselves?
Search Console now carries a generative AI performance report for Google's AI features. For non-Google engines there is no first-party report, so you are relying on third-party indexes, which is exactly why we print the sample size and the limits beside every number here.
Sources
Adith Krishnan
Co-Founder & COO, Kula Digital
8+ years building marketing that is measured in revenue. Runs strategy, AI search visibility, and education marketing at Kula. Written from the studio in Coimbatore.