Capability

AI visibility: be findable when buyers ask an assistant

Buyers increasingly ask ChatGPT, Perplexity or an AI Overview instead of scanning ten blue links. PolyDraft ships the concrete artifacts that make a site accessible and quotable to those systems — llms.txt, an AI-crawler-friendly robots.txt, and answer-first structure — and is explicit about what nobody can promise.

Definition

What AI visibility means

AI visibility is how findable and quotable a website is when someone asks an AI assistant instead of running a search. It rests on three conditions: AI crawlers can reach your pages, your site publishes a machine-readable map of what it covers, and each section still carries its meaning when lifted out of the page around it.

The practical shift is the unit of optimization. Search engines rank pages; assistants retrieve passages. A page that reads beautifully in sequence but whose fourth section is meaningless without the first three is a page an assistant cannot safely quote — so it quotes a competitor who structured theirs to stand alone.

What ships

Artifacts, not adjectives

Most builders describe AI readiness as a feeling. These are files and structures you can fetch and inspect on your own site the moment it is built.

llms.txt on every site

A curated Markdown map of your key pages at /llms.txt, following the public proposal from llmstxt.org. It gives an assistant a compact, accurate summary of what your site covers instead of leaving it to infer structure from HTML.

robots.txt that welcomes AI crawlers

GPTBot, ClaudeBot and PerplexityBot are named and allowed explicitly. Many sites block them by accident through a restrictive default; being absent from the crawl is the one failure mode that guarantees you are never quoted.

Answer-first section structure

Key sections open with the claim and follow with support, so a section carries its meaning when lifted out of the page. Assistants retrieve passages, not documents — a section that only makes sense in sequence is a section that cannot be quoted.

Consistent entity naming

One name for each product, service and brand across the entire site. Synonyms and abbreviations fragment the entity a retrieval system builds of you.

Schema.org structured data

Machine-readable statements of what each page is about, who published it and how the site is organized — the same graph that powers search rich results.

A blog feed and honest lastmod

RSS plus per-page modification dates that only change when the page changed, so freshness signals mean something to anything that consumes them.

Build either file for any site with the free llms.txt generator and robots.txt generator. The underlying discipline is covered in generative engine optimization.

The agents

Who is actually reading your site

Each operator documents its own crawler and its own controls. Links go to the official source, because these policies change and second-hand summaries go stale.

AI crawlers welcomed by default in a generated robots.txt, with their operators and official documentation
Crawler Operator What it does
GPTBot OpenAI Crawls public pages that may inform ChatGPT responses.
ClaudeBot Anthropic Anthropic’s web crawler; site owners control access via robots.txt.
PerplexityBot Perplexity Crawls pages that can be surfaced and cited in Perplexity answers.
Google-Extended Google A robots.txt control token governing Gemini and Vertex AI grounding use.

Allowing these crawlers is the default, not a lock-in: blocking any of them from the admin panel is a legitimate choice if you would rather your content stay out of AI systems.

The workflow

How the layer is applied

01

Make the site crawlable by the right agents

robots.txt is generated with AI crawlers explicitly allowed, and the sitemap lists every canonical URL. Access is the precondition for everything else on this page.

02

Publish a machine-readable map

llms.txt is generated from your actual page set and refreshed as pages change, so the summary an assistant reads never describes a site you no longer have.

03

Structure sections to survive extraction

Content is written answer-first and split into self-contained sections, with FAQ blocks marked up so the visible text and the structured data match verbatim.

04

Keep the entity consistent as you publish

New pages and posts inherit the same naming, the same structured-data graph and the same internal-link dictionary, so the picture does not fragment as the site grows.

The recurring technical work underneath it is described on SEO automation.

Limits

What we do not promise

This is the part of the category most prone to overselling, so here is the honest boundary.

No vendor guarantees citation, and neither do we

No AI provider publishes a mechanism that turns a well-structured site into a guaranteed citation. What we ship is the input side: access, a clean map, and extractable structure. Whether an assistant quotes you is its decision, not a purchasable outcome.

llms.txt is a proposal, not a ratified standard

It is a public specification with growing adoption, not an IETF standard, and support varies between providers. We publish it because the cost is near zero and the file is genuinely useful to humans too — not because anyone has proven it changes rankings.

Crawler behaviour changes without notice

User-agent names, controls and policies are set by each provider and can change at any time. We track the official documentation linked on this page rather than making claims that quietly go stale.

Visibility still needs something worth quoting

Structure makes content extractable; it does not make thin content valuable. Pages with real operational specifics get quoted. Pages that restate what every competitor says do not, however well marked up.

Standards and official documentation

Primary sources

Read these directly rather than trusting anyone’s summary, including ours — the policies here change more often than in classic search.

AI visibility questions

What is AI visibility?

AI visibility is how findable and quotable your website is when someone asks an AI assistant instead of typing a search query. It depends on three things: whether AI crawlers can access your pages, whether your site publishes a machine-readable map of what it covers, and whether individual sections still make sense when extracted from the page around them.

Is AI visibility the same as SEO?

It extends SEO rather than replacing it. Classic SEO optimizes a page so it ranks in a result list. AI visibility optimizes a section so it can be retrieved and quoted inside an answer. The technical fundamentals overlap almost entirely — crawlability, structured data, clear headings — which is why a well-built site starts from a strong position.

Does publishing llms.txt make an AI cite my site?

No one has demonstrated that, and we do not claim it. llms.txt is a public proposal that gives an assistant a compact, accurate map of your pages; support varies by provider and it is not a ratified standard. We publish it because it costs almost nothing, is useful to human readers too, and removes ambiguity about what your site covers.

Which AI crawlers does a PolyDraft site allow?

The generated robots.txt explicitly welcomes GPTBot, ClaudeBot and PerplexityBot alongside classic search crawlers. You can change this at any time from the admin panel — the file is yours, and blocking any of them is a legitimate choice if you prefer your content stay out of AI training or grounding.

How would I know whether it is working?

Two signals, on different timescales. Immediately, you can verify the mechanics: fetch your llms.txt and robots.txt, confirm your structured data parses, and check that individual sections read correctly in isolation. Over weeks, you can watch for referral traffic from assistant surfaces and try your own buying questions in ChatGPT or Perplexity to see whether you appear.

Check your own site first

The free tools need no account: generate an llms.txt and a robots.txt for the site you have today, then build a full PolyDraft site free and compare what ships by default.