AI-led Services

Buyers now ask ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews the questions they used to type into a search box. Whether your business shows up in those answers depends on things most marketing teams have never looked at: which AI crawlers your robots policy lets in, whether your pages answer a question in the first paragraph, whether your entities and structured data are unambiguous, and whether you can even see AI assistant traffic in your analytics. This engagement covers all of it, and it is the same playbook running on this site: AI crawler logging, an AI discovery report, llms.txt, and an MCP server listed in the official MCP Registry.

Book an AI strategy call All ai services services

What it moves

  • 6Workstreams: crawler access, llms.txt, content, entities, measurement and agent readiness
  • 1Playbook, already running on rahuldsarker.co before it is sold to anyone else
  • 0Made-up visibility scores: measurement starts from your own server logs and analytics
  • 0Ranking guarantees: nobody controls what an AI assistant says, and we will not pretend to

Why most businesses are invisible to AI assistants without knowing it.

AI search is not a separate channel with its own tricks. It is a set of crawlers, indexes and answer engines that read your site differently from Google. The failure points are technical, editorial and measurement problems, and most of them are quick to find once someone looks.

  • The robots policy blocks the crawlers that matter, or allows the ones you did not intend.

    A blanket rule copied from a forum blocks every AI user agent, including the search and live-answer crawlers that decide whether you can be cited. Or the opposite: everything is open, including training crawlers the business would rather opt out of. Nobody has separated training, AI search indexing and live answer fetching, so the policy is an accident rather than a decision.

  • The content buries the answer.

    The page opens with three paragraphs of brand story before it says what the service is, who it is for and what it costs to scope. AI assistants extract passages that answer a question directly. A page that never states its answer plainly gives them nothing to quote, so they quote a competitor or a directory listing instead.

  • Your entities are ambiguous.

    The company name, the founder, the services and the locations are described differently on the website, LinkedIn, directories and review sites. There is no Organization or Person schema tying them together, no consistent sameAs links, and no single page that states the facts. An AI system that cannot resolve who you are will not confidently recommend you.

  • AI assistant traffic is hidden inside "referral" and "direct".

    Visitors arriving from ChatGPT, Perplexity, Copilot or Gemini land in analytics as generic referral or direct traffic. Crawler visits never show up in client-side analytics at all, because crawlers do not run JavaScript. The business cannot answer the basic question of whether AI discovery is happening, so it cannot decide whether to invest in it.

  • Nothing is ready for AI agents.

    Assistants are starting to act, not just answer: fetching pages on a user's behalf, calling tools and comparing vendors. There is no llms.txt, no clean machine-readable summary of what the business does, and no thought given to whether a tool or MCP server would make the business easier for an agent to work with.

How GEO and AI search optimisation works.

The engagement starts with access and measurement, because there is no point rewriting content for crawlers that are blocked, or optimising for traffic you cannot see. Content and entity work follows, then agent readiness.

  1. Phase 1

    AI crawler access and robots policy

    • Robots.txt and firewall audit, which AI user agents are allowed, blocked or silently challenged by your CDN or bot protection
    • Crawler purpose map, training, AI search index and live answer crawlers separated so each gets a deliberate allow or disallow decision
    • Policy recommendation, written rules for which crawlers to allow and why, agreed with you before anything changes
    • Rendering check, whether your key pages deliver their content in the HTML or only after JavaScript runs
    • Sitemap and canonical review, so crawlers find the pages you want cited rather than duplicates and parameters

  2. Phase 2

    Measurement: AI assistant traffic and crawler activity

    • Server-side AI crawler logging, crawler visits identified by user agent and logged at the edge or server, since they never appear in client-side analytics
    • AI referral channel grouping, visits from ChatGPT, Perplexity, Claude, Gemini, Copilot and similar assistants split out from generic referral traffic
    • AI discovery report, which pages AI crawlers fetch, which assistants send visitors, and which pages those visitors land on
    • Baseline snapshot, the starting position recorded so later changes can be compared against your own data rather than a vendor score
    • Caveats documented, user agents can be spoofed and some assistants pass no referrer, so the report states what it can and cannot see

  3. Phase 3

    Answer-first content and entity clarity

    • Priority page rewrite plan, service, pricing-scope, comparison and FAQ pages restructured to state the answer in the opening lines
    • Question mapping, the questions buyers actually ask assistants about your category, matched to the page that should answer each one
    • Entity consolidation, one consistent description of the business, founder, services and locations across site and key profiles
    • Structured data, Organization, Person, Service, FAQPage and BreadcrumbList schema implemented and validated, tied together with stable @id references
    • llms.txt, a curated, plain-language map of your most important pages for language models, kept in step with the sitemap

  4. Phase 4

    Agent readiness and handover

    • Agent readiness review, whether an AI agent acting for a buyer can find your offer, understand your scope and reach you without friction
    • MCP and tool assessment, whether exposing a tool or MCP server (for example a calculator or product lookup) is worth building for your business
    • MCP build and registry listing where it makes sense, scoped separately, following the same route used for this site's own MCP server
    • Monitoring cadence, a monthly review of the AI discovery report and crawler logs with a short list of next actions
    • Documentation and handover, robots policy, schema map, llms.txt process and reporting explained so your team can maintain them

What is included in GEO and AI search optimisation.

Access

  • AI crawler robots policy
  • CDN and bot protection review
  • Rendering check
  • Sitemap and canonical review
  • Written allow and block rationale

Measurement

  • Server-side AI crawler logging
  • AI assistant referral grouping
  • AI discovery report
  • Baseline snapshot
  • Monthly review template

Content and entities

  • Answer-first page rewrites
  • Buyer question map
  • Entity consolidation
  • Schema implementation
  • llms.txt

Agents and handover

  • Agent readiness review
  • MCP and tool assessment
  • Optional MCP build and registry listing
  • Maintenance documentation
  • Team walkthrough

This is right for you if:

  • B2B companies and consultancies whose buyers research vendors by asking AI assistants before they ever fill in a form
  • Marketing leaders who suspect AI discovery is happening but cannot see it in their analytics
  • Teams with solid SEO foundations who want to extend them to AI crawlers, answer engines and agents rather than start a separate project
  • Founders who want a practitioner who has implemented crawler logging, llms.txt and an MCP server on a live site, not a slide deck on "AI visibility"

Not the right fit if:

  • Businesses looking for guaranteed placement in ChatGPT or Google AI Overviews: nobody can honestly promise that
  • Sites with no indexable content yet, the foundations of a crawlable, useful website have to exist before AI search optimisation can help

Frequently asked questions.

What is GEO (generative engine optimisation)?

GEO is the work of making a website easy for AI systems to crawl, understand and cite. In practice it covers four things: letting the right AI crawlers in through a deliberate robots policy, structuring content so it answers questions directly, making your entities and structured data unambiguous, and measuring AI crawler activity and AI assistant referrals so you know whether any of it is working. It builds on SEO rather than replacing it.

Should we block AI crawlers or allow them?

It depends on the crawler's purpose. AI companies run separate user agents for model training, for building an AI search index, and for fetching pages live when a user asks a question. Many businesses choose to allow search and live-answer crawlers, because those decide whether they can be cited, while making a separate decision on training crawlers. The engagement maps each crawler to its purpose and documents the decision rather than applying a blanket rule.

What is llms.txt and does it matter?

llms.txt is a proposed convention: a plain-text file at the root of your site that gives language models a curated summary of who you are and links to your most important pages. It is not an official standard and no AI company guarantees it is used, so it is a low-cost addition rather than a strategy on its own. We set it up, keep it in step with your sitemap, and put most of the effort into the content and access work that matters more.

How do you measure AI search visibility?

From your own data rather than a third-party score. AI crawler visits are logged server-side by user agent, because crawlers do not run analytics scripts. AI assistant referrals are separated into their own channel in analytics. Together they show which pages AI systems fetch and which assistants send real visitors. User agents can be spoofed and some assistants send no referrer, so the report states its blind spots plainly.

Do we need an MCP server?

Most businesses do not, at least not yet. An MCP server lets AI assistants call a tool you provide, such as a calculator, a product lookup or a booking check. It is worth building when you have something genuinely useful for an agent to do. This site runs its own MCP server for marketing calculators, listed in the official MCP Registry, so the assessment is based on having built and listed one, not on theory.

How is the cost scoped?

By the size of the site and how many workstreams you need. The main variables are the number of priority pages to restructure, how complex your CDN and bot protection setup is, whether server-side crawler logging can be added to your hosting, and whether an MCP build is in scope. After a 30-minute call we send a fixed-scope proposal with each workstream priced separately, so you can start with access and measurement and add the rest later.

Want to know whether AI assistants can find and cite your business?

Book a 30-minute call. We will look at your robots policy, a few priority pages and what your analytics can currently see, and tell you which workstream to start with.

It starts with a 30-minute call. Pick a time below.

Choose a time

Prefer email?