Getting quoted when the answer engine replies

What we're honest about before you pay for this

There is no published, independently verifiable study that tells anyone exactly how ChatGPT, Perplexity or Google's AI Overviews choose which source to quote. Anyone who shows you a chart claiming to have cracked the ranking factors of a large language model has made the chart up.

What can be said confidently is mechanical. These systems retrieve documents, then generate an answer from passages inside them. A passage that already reads as a complete, attributable answer is easier to lift than one that needs three paragraphs of context to make sense. Structured data is easier to parse than an implied claim. A brand that resolves to a consistent entity across the web is easier to name than one spelled four different ways.

That's the basis for the work. It's engineering against how retrieval and generation function, not a decoded algorithm. If you'd rather see the argument in full, we've written about whether generative engine optimisation is real without the pitch attached.

The answer-block pattern we apply to every page

The core edit is small and it changes almost every page on a site. Under each H2, phrased as the question a person would actually type, we place a direct answer of roughly 40 to 60 words that makes complete sense with zero surrounding context. Everything else — the nuance, the caveats, the examples — comes after it.

Forty to sixty words is the target for a specific reason: it's long enough to contain a claim, a qualifier and a number, and short enough that a model can drop it into an answer without editing it. Below thirty words you're usually asserting without supporting. Above eighty, the model starts summarising you, and summarised sources get cited less consistently than quoted ones.

• Question-shaped headings. 'How much does SEO cost in India' beats 'Pricing'. The heading is a retrieval hook, not a design element.
• Self-contained answers. No 'as we mentioned above', no pronouns pointing at the previous paragraph. Assume the block is read alone, because it will be.
• A number, a range or a named thing in every block. Specifics survive summarisation. Adjectives don't.
• Dates on claims that decay. 'As of July 2026' gives a model a reason to prefer you over an undated page from 2021.
• One idea per block. Two claims fused into one paragraph get split, and the half that gets quoted is rarely the half you cared about.

Entity work: making a model resolve you as a named source

A model can only cite something it can name. If your company appears as 'Last Agency', 'LastAgency' and 'Last Agency Pvt Ltd' across LinkedIn, your own footer and every directory listing, you're three weak entities instead of one strong one.

Entity work is the least visible part of this service and the part that compounds. It's also cheap, because most of it is a one-time cleanup rather than an ongoing programme.

• `Organization` schema with a complete `sameAs` array pointing at your LinkedIn, Crunchbase, GitHub and any profile you control. This is the machine-readable statement that all those profiles are one company.
• One canonical name and one description, used identically everywhere. Boring, and it's the hour in this engagement that pays back hardest.
• Real authors with real credentials, marked up as Person, linked to their own profiles. Anonymous content is harder to attribute and easier to ignore.
• Unlinked brand mentions, which count for entity resolution even where the link doesn't exist. This is why digital PR matters here as much as it does for classic rankings.
• Third-party corroboration — a Wikidata entry if you genuinely qualify, listings on the platforms your category uses, consistent details on every one of them.
• `FAQPage` and `Article` markup. Google stopped showing FAQ rich results for most sites in 2023, so this isn't a snippet play any more. We still mark it up because it's cheap, unambiguous structure for anything parsing the page.

How we measure citation share, and what the number means

Assistants are non-deterministic. Ask the same question twice and you can get two different sets of sources. Ask from Mumbai and from London and the answers diverge again. Any agency reporting 'you are cited for X' from a single check is reporting a coin flip.

So we measure a sample, not an instance. At kickoff we agree a frozen panel of prompts — typically 40 to 80, written the way your buyers actually phrase things, covering the full funnel from category education to 'best X in India'. That panel doesn't change during the engagement, because changing the questions is how agencies manufacture improvement.

1. Run the frozen panel monthly across ChatGPT, Perplexity and Google AI Overviews, at a consistent time and from consistent locations.
2. Log three things per prompt: whether you were mentioned, whether you were linked, and which specific URL got the citation.
3. Report citation share — the percentage of panel prompts where you appear — plus which competitors appear most often and which of your pages are doing the work.
4. Trace the winners. When a page starts getting cited, we look at what changed and apply the same treatment to the pages that aren't.
5. Report the misses too. Prompts where a competitor is consistently quoted and you aren't are the most useful lines in the report, and they're the ones most reports omit.

Where this stops and AI search optimisation starts

We split AI visibility into two services because they're genuinely two different jobs, done by different people, and bundling them lets agencies charge twice for whichever half they were going to do anyway.

This page is content and structure. The other half — AI search optimisation — is access and plumbing, proving the crawlers can actually reach you. If an assistant is blocked at your CDN, no amount of answer-block rewriting will help, which is why we usually run the access audit first.

What's included, what it costs, and what the guarantee covers

Answer engine optimisation starts at ₹75,000 a month, ex-GST, month-to-month after the first quarter, thirty days' notice, and you keep every asset. Smaller sites with fewer than about 30 pages start at ₹40,000 a month. See pricing for the rest, including the bundle if you're running SEO and social with us.