AEO Services (Answer Engine Optimisation)

What an answer engine actually does before it cites you

Almost every AI answer you see on the public web is retrieval-augmented. The engine turns the user's question into one or more search queries, pulls a candidate set of pages from an index, re-ranks that set, reads the top few, then writes an answer and attaches citations to the pages it leaned on. Nothing in that chain is magic, and none of it happens if your page is missing from the index, blocked to the retrieval crawler, or written so that no single paragraph answers the question on its own. AEO is therefore three jobs stacked: be crawlable, be in the index, be extractable. Most sites fail at one of the first two and never find out.

Crawler access: the gate nobody checks

AI companies run separate agents for separate jobs, and blocking one does not block the others. OpenAI runs GPTBot for model training, OAI-SearchBot for the index behind ChatGPT search, and ChatGPT-User for fetches a user triggers. Anthropic runs ClaudeBot alongside its user-triggered and search agents. Perplexity runs PerplexityBot for indexing and Perplexity-User for live fetches. Google-Extended is not a crawler at all — it is a robots.txt product token that governs whether content Googlebot already fetched may be used for Gemini Apps grounding and model training. We read your robots.txt, your CDN and WAF rules, and your raw server logs, because plenty of sites block these agents at the firewall without knowing it.

Writing passages a model can lift without breaking them

Retrieval works on passages, not on whole pages. A passage that survives extraction states the subject by name rather than as "we" or "it", answers in the first two sentences, carries its own dates and figures, and does not depend on the paragraph above it for meaning. Tables, short definition blocks and question-shaped headings all extract cleanly. Long build-up paragraphs, testimonial padding and pages that cover four topics at once do not. We rewrite your key pages so that every question you want to own has one clean, self-contained block answering it, placed high on a page that is genuinely about that question.

Schema, llms.txt, and honest expectations about both

Structured data does not force a citation. What it does is remove ambiguity: Organization and Person schema tied to a single stable @id, Article with a real named author, FAQPage for genuine question-and-answer blocks, and Breadcrumb for context. Note that Google restricted FAQ rich results to a narrow set of sites, so treat FAQPage as machine-readable clarity rather than a SERP feature. Speakable markup is a limited, news-focused feature — we add it only where it applies. On llms.txt: it is a proposed convention, no major AI provider has confirmed using it, and Google has said it has no effect on Search. It costs an hour to publish, so we publish it, and we do not sell it as a lever.

Why Bing still matters, and what we can and cannot promise

Microsoft Copilot grounds its answers in Bing's index, so absence from Bing means Copilot cannot cite you. ChatGPT has historically drawn on Bing results and now also runs its own search index, which makes Bing Webmaster Tools verification, a submitted sitemap and IndexNow pings cheap insurance rather than a fashion. What we will not do is promise a citation. Nobody controls what a model writes on any given day, exactly as nobody controls a Google ranking. The difference worth paying for is eligibility: most sites are not even in the candidate set, and that part is diagnosable and fixable. Audit ₹10,000 one-time; technical SEO ₹25,000/month; on-page ₹15,000/month.

AEO Services (Answer Engine Optimisation) — FAQs