Generative Engine Optimisation, without the parts nobody can deliver
Answer engine optimisation services in India, built on what OpenAI, Google, Anthropic and Perplexity actually document rather than on what the category is currently selling. Half of this page is the list of things we will not promise you.
What nobody can sell you, us included
This category is eighteen months old and already full of promises that the engine operators' own documentation contradicts. Every line below was checked against the primary source in September 2026. If a proposal you are reading contains one of these, the rest of it is worth less than it looks.
No operator offers an inclusion guarantee, and Google's own guidance on hiring an SEO says plainly that no one can guarantee a #1 ranking and to beware of anyone alleging a special relationship with Google.
There is no shared result page for a position to exist on. ChatGPT personalises answers, so two people asking the same question can get different sources. A number here is not hard to achieve, it is incoherent.
There is no submission endpoint or indexing API for ChatGPT, Claude or Perplexity. OpenAI documents four crawlers and no submit form. Anyone selling submission is selling something that does not exist.
Google states Search does not use the file. Ahrefs looked at 137,210 domains and found 97% of published llms.txt files received zero bot requests in the month measured. We publish one and we still will not sell it to you as a lever.
A controlled Ahrefs study tracked 1,885 pages that added JSON-LD against roughly 4,000 matched controls and found essentially no effect on AI citations. Schema earns rich results in ordinary Search. That is a good enough reason to do it.
LLM answers are non-deterministic, so a tracked position is a sample statistic rather than a rank. A chart with Search Console semantics implies a stability that is not there. We will show you sampling, with the method attached.
What actually decides whether an AI answer can quote you
Google's documented bar for appearing in AI Overviews or AI Mode is exactly this: “a page must be indexed and eligible to be shown in Google Search with a snippet.” If nosnippet or noindex is set, the page is out. That is the whole technical requirement, and it is ordinary SEO.
Verbatim from Google: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.” Any GEO package built on secret markup is charging you for a file the largest AI surface in India has said it does not read.
Allowing GPTBot is a training decision. The agent OpenAI documents for ChatGPT search is OAI-SearchBot, where an opted-out site “will not be shown in ChatGPT search answers.” Anthropic splits training from Claude-SearchBot the same way. Perplexity is the opposite case: PerplexityBot is a search bot and is “not used to crawl content for AI foundation models.” Three operators, three different meanings, and one robots.txt file that has to get all three right.
An answer engine lifts a specific sentence. Prices stated as numbers, processes stated as steps, and questions answered in the first line get quoted. Copy that circles the point for two paragraphs before committing does not, whatever the schema says.
We got this wrong on our own site first
In September 2026 we published an llms.txt for this site and wrote that it would help AI engines cite us. Then we read Google's documentation properly, which says Search does not use the file, and an Ahrefs study of 137,210 domains showing 97% of such files receive no bot requests at all. The file is still there, because it is accurate and costs nothing. The claim is gone.
The same week we found our own robots.txt naming GPTBot, ClaudeBot and Google-Extended, which are training and grounding agents, while the agents that govern search visibility were unnamed. Worse, each named group silently replaced the site-wide disallow rules, so the AI crawlers were the only ones we were letting into paths everything else is kept out of.
That is the work. Not a secret file, not a submission, not a ranking chart. Reading three sets of operator documentation that disagree with each other, checking what your site actually sends, and fixing the gap between them. See SEO services for the technical work this sits on top of.
Questions we get asked
What is generative engine optimisation, in plain terms?
Work that makes a page easy for an AI answer to find, read and quote correctly. In practice most of it is ordinary technical SEO done properly, because Google's documented requirement for AI Overviews is that a page is indexed and snippet-eligible, plus crawler access decisions that differ per operator, plus writing that states facts a machine can lift.
Are answer engine optimisation, GEO and ChatGPT SEO different services?
No, and we deliberately have one page for all three rather than three pages competing with each other. The vocabulary is genuinely unsettled. People search for answer engine optimization services in India and for generative engine optimization agency in India, usually with the American spelling because that is how the industry writes it, while Indian business writing tends to use optimisation. Others search for ChatGPT SEO services in India, or AEO, or GEO. Different words, same work underneath, and we would rather answer to all of them than pretend they are separate products with separate price lists.
Can you guarantee we get cited by ChatGPT or in an AI Overview?
No, and neither can anyone else. There is no submission endpoint, no inclusion guarantee from any operator, and no published ranking factors. What we can do is remove the reasons a page gets skipped and make it quotable, then measure honestly. If an agency offers you a guarantee here, that alone tells you what their measurement will be worth.
Then what do you actually measure?
Search Console's generative AI reporting for impressions, which is first-party and real, with the caveat that it groups AI Overviews and AI Mode and does not give per-surface clicks. Referral traffic from AI assistants in analytics, which is small but genuine. And prompt sampling with the method written down: how many prompts, how many runs each, which engines, and the run-to-run variance. A single AI visibility score with none of that attached is a number for a slide.
Do we need this if our SEO is already being done properly?
Often not, and we would rather say so. If your pages are indexed, snippet-eligible, fast and clearly written, most of the work is already done. The genuine gaps we find tend to be crawler directives that say the opposite of what the owner intended, and content that never states a fact plainly enough to be quoted.
You publish an llms.txt. Does that not contradict this page?
It would if we sold it to you. We publish /llms.txt and /llms-full.txt because they cost one generated file per build and they are accurate, and some systems do read them. Google states Search does not, and the largest study to date found 97% of such files get no requests at all. Keep one if you like. Do not buy one.
Who is this for?
Businesses whose buyers ask assistants questions your website should be the answer to, usually because the purchase involves research: software, services, healthcare, education, B2B products. If your customers find you through a map and a phone call, your money is better spent on the Google Business Profile.
Start with the audit
We will tell you what your site currently tells each crawler, whether your pages are snippet-eligible, and which of your content an answer engine could actually quote. If the answer is that your SEO already covers it, we will say that too.