How AI sees it: invisible · preview · 21 critical issues blocking AI agents.
AISO fixAllowlist these UAs at your CDN (Cloudflare WAF rules / Akamai Bot Manager / fastly). Cloudflare's 'Verified Bots' toggle covers most of these but ChatGPT-User and PerplexityBot frequently need explicit rules. Check robots.txt too.
AISO fixMaintain readable structure + recency: clear headings/lists for chunkability, recent updated-at dates, no stale fact decay, answer-extractable formatting (Q→A, def→explanation).
AISO fixCite at least one .edu/.gov, Wikipedia, IETF/W3C spec, or tier-1 publication near your flagship claims.
AISO fixWrap the FAQ with a FAQPage JSON-LD block so AI engines can lift the Q/A pairs verbatim.
AISO fixMake WHO you are unambiguous: Organization JSON-LD with sameAs links, agent-discovery files (llms.txt + .well-known manifests), structured contact + offer data so delegate agents can transact.
AISO fixMost claims lack inline sources within 200 chars. LLMs heavily discount uncited statistics. Add anchor links to primary sources, "according to" attributions, or numeric footnote refs.
AISO fixAdd headings like '<Product> vs <Competitor>' to comparison content
AISO fixDeclare a PotentialAction (e.g., BuyAction, BookAction, OrderAction) with target endpoint + input schema so delegate agents know how to invoke the transaction.
AISO fixAdd a /best-X-alternatives or /you-vs-competitor page
AISO fixOpen the homepage with '<Brand> is a <category> that <differentiator>' — this is the single biggest lever for AI citability.
AISO fixAdd FAQPage schema with 5-10 high-frequency questions for AI-search retrieval.
AISO fixAdd an HTML <table> with <th>feature</th>/<th>price</th> columns
AISO fixAdd a glossary page with 5+ defined terms
AISO fixDefine key terms explicitly: '**AI Agent** is a software program that...'
AISO fixAdd Pros/Cons sections to comparison and review pages
AISO fixAdd an explicit '<h2>FAQ</h2>' or 'Frequently Asked Questions' section with at least 3 items.
AISO fixAdd Review and/or AggregateRating JSON-LD with reviewer + ratingValue + datePublished. LLMs cite review summaries when ranking peers in comparison answers.
AISO fixscript-src 'self'; default-src 'self';
AISO fixIncrease the sampled-page count to evaluate internal connectivity
AISO fixEvery % / 'X customers' / stat deserves an inline citation link to the origin study or publication.
AISO fixValidate the response from your manifest URL. The body must parse as JSON. Re-serve with `Content-Type: application/json`.
AISO fixReturn valid JSON at /.well-known/agent.json describing your agent or MCP capabilities.
AISO fixReturn valid JSON at /ai/faq.json with a faqs array of {question, answer, source_url} objects.
AISO fixReturn valid JSON at /ai/service.json shaped like a schema.org Service (@type, name, provider).
AISO fixReturn valid JSON at /ai/summary.json with at least name and description fields. Validate with a JSON linter.
AISO fixAllowlist these UAs at your CDN (Cloudflare WAF rules / Akamai Bot Manager / fastly). Cloudflare's 'Verified Bots' toggle covers most of these but ChatGPT-User and PerplexityBot frequently need explicit rules. Check robots.txt too.
AISO fixBreak paragraphs above 400 words with subheadings, lists, or bolded facts
AISO fixPrioritize: Frequently Asked Questions about AI Agent Tooling; AI Agent Tooling Glossary: Key Terms Explained.
AISO fixLift the lowest-scoring dimensions in `composite_inputs` first; citation-crawl and agent-readiness move the headline metric the most.
AISO fixEmit Article JSON-LD with headline, author, datePublished, dateModified.
AISO fixbase-uri 'none'; (or 'self' if relative URL rewriting is required)
AISO fixAdd an `audience` property (or Audience node) to your Organization/Service JSON-LD with `audienceType` (e.g. 'Developers', 'Marketers'). LLMs answering 'who is X for?' use this directly.
AISO fixAdd a 2–4 sentence paragraph immediately after the H1 summarising what the product is and who it serves.
AISO fixPublish an AGENTS.md at your repo or web root listing the skills, conventions, and tools agents should use when working with your codebase or service.
AISO fixOpen with 'We are ...' / 'Our mission is ...' / 'This is ...' — AI models need explicit entity framing.
AISO fixAdd a ContactPoint node with telephone or email, ideally including contactType. Agents fall back to noisy prose extraction without this.
AISO fixAdd a TL;DR sentence or '<Brand> is a <category> that …' lead at the top
AISO fixAdd a Person JSON-LD entry for founder/principal author with sameAs links.
AISO fixAdd <time datetime='YYYY-MM-DD'>Updated ...</time> or article:modified_time meta, or expose dateModified in Article JSON-LD.
AISO fixPublish an MCP endpoint or setup guide, then link it from the homepage or llms.txt. Agents should be able to discover tool access without reading human-only docs.
AISO fixAdd a PostalAddress node (or address property on Organization) with addressLocality + addressRegion/Country so delegate agents can geo-resolve the business.
AISO fixobject-src 'none';
AISO fixAim for 134-167 words per h2-bounded section — zubair's empirical citability optimum. Long sections get truncated; short ones don't carry enough signal.
AISO fixRestructure paragraphs to 30-120 words each. RAG retrievers favour self-contained spans of this size.
AISO fixAdd 'according to', 'source:', or numeric footnote references. LLMs cite content that points at primary sources.
AISO fixDelay or remove modal and overlay patterns on indexable pages, especially above-the-fold newsletter or signup gates.
AISO fixRemove agent-hostile signals: prompt-injection patterns, broken links, deceptive cloaking, robots conflicts. Agents discount hosts that look unsafe to act on.
AISO fixUse exactly one H1, ensure the first H2 precedes any H3, and give every H2 a clear topic label (3+ words).
AISO fixAdd statistical claims to drive AI-citation rate +30-40% (Aggarwal Princeton GEO 2024).
AISO fixAim for 3-11 words of context-bearing description per non-decorative image.
AISO fixAdd explicit entity surfaces: team or founder page, pricing page, docs or API page.
AISO fixLong sentences degrade embedding quality. Tighten sentences to ≤ 25 words for clean chunk embeddings.
AISO fixdefault-src 'self';
AISO fixImprove the underlying dimensions that drive google_ai citations: crawlability, structure, schema, freshness, entity clarity, and trust.
AISO fixAdd ItemList JSON-LD schema to comparison pages
AISO fixWrap term/definition pairs in <dl><dt>term</dt><dd>def</dd></dl> markup
AISO fixIf the homepage surfaces editorial content, include an author byline (Person JSON-LD or 'By <Name>').
AISO fixIf your business has operating hours, declare openingHoursSpecification so agents only attempt transactions during open windows.
AISO fixDeclare paymentAccepted on your Offer/Service so payment-capable agents can confirm method compatibility before initiating a transaction.
AISO fixIf the service supports temporary agent visits, publish a Portal manifest at /portal with tool metadata and a call endpoint, then link it from docs and llms.txt.
AISO fixAdd `serviceType` (e.g. 'AI search optimization') or `suitableForBusinessFunction` to your Service/Product node so LLMs categorise the offering correctly.
AISO fixIf the site supports browser-native agent access, publish a WebMCP endpoint or discovery document and link it from developer docs and llms.txt.
AISO fixIf paid agent/API access is supported, document x402 payment flow, supported networks/tokens, payment headers, and facilitator behavior.
AISO fixAdd 'founder': [{ @type: Person, name: ... }] to the Organization schema.
AISO fixWe will retry on the next scan. No action needed on your side.
AISO fixImprove the underlying dimensions that drive perplexity citations: crawlability, structure, schema, freshness, entity clarity, and trust.
AISO fixIf you launch localised variants (e.g. /es, /de, fr.example.com), add `<link rel="alternate" hreflang="...">` tags pointing at each variant plus `x-default`.
AISO fixRun Template Audit, Full Site, or Deep Investigation mode to compare key templates across the site.
AISO fixupgrade-insecure-requests;
AISO fixPages about prices, versions, launches, or changing data should expose a clear last-updated date and refresh cadence.
AISO fixAdd page-family schema for Article/BlogPosting/TechArticle, BreadcrumbList, Person. Strong JSON-LD should match the page type, not just include generic Organization markup.
Deterministic extraction from stored page context: identity, offers, actions, files, trust, and evidence.
No grounded description found.
No structured offer or price evidence found.
No booking, buying, contact, quote, or API path found.
No llms.txt, agents.json, mcp.json, x402.json, profile.json, or offers.json found.
Agents need a stable name, category, and short description before they can explain the page.
Agents cannot recommend, compare, or transact without a grounded offer or price.
Agents need explicit next actions instead of guessing how to engage.
Files such as llms.txt, agents.json, mcp.json, profile.json, and offers.json make the site directly readable.
Paste these into the public agent files your site is missing, then rescan.
# warp.dev
> Public profile for agent-readable discovery.
Canonical: https://warp.dev/
Agent-readable schema: aios.agent_readable.v1
## Offers
- Add structured offers.
## Actions
- Add book, buy, quote, contact, or API actions.{
"schema": "aios.agent_readable.v1",
"name": "warp.dev",
"url": "https://warp.dev/",
"description": "Public profile for agent-readable discovery.",
"offers": [],
"actions": [],
"evidence": []
}{
"schema": "aios.mcp_manifest.v1",
"name": "warp.dev",
"profile_url": "https://warp.dev/",
"tools": [
{
"name": "ask_profile",
"description": "Answer questions using grounded public profile context."
},
{
"name": "get_offers",
"description": "Return structured offers with prices, URLs, and evidence."
},
{
"name": "start_action",
"description": "Open the best booking, buying, quote, contact, or API action."
}
]
}Every check rolls up into covered, not captured, or planned-only. Click-through rows below stay tied to stored signals only.
AISO maps llms.txt and WebMCP overlap with Chrome Lighthouse's experimental agentic audits. This is an overlap map, not a Lighthouse pass/fail claim.
Web3 surfaces, Solidity and Rust audit, content quality, and site-wide context intelligence ship via the AISO Web Context Engine plus planned TOOLBOX code-audit workers. Cards below show what each module checks; metrics populate once the context-shard dispatch is enabled for your tier.
Wallet-connect, contract addresses, x402 endpoint health, ENS, chain metadata, exposed ABI and token metadata.
Context Engine - scrape regex + actionsFlesch-Kincaid readability, AI-readability, thin pages, duplicate content, keyword stuffing and original-research density.
Context Engine - scrape + researchSitemap coverage %, scrape markdown quality, crawl-discovered pages, entity knowledge-graph presence, research-synthesized authority and wire-extractor results.
Context Engine - scrape + map + crawl + researchAgent readiness is discovery, metadata, proof, offers, actions, protocols, and safe access together.
Sequential. Most of it auto-generates when you install Agent Link; the scanner verifies the result on the next pass.
The preview above stays visible. The detailed scanner rows, provider evidence, PDF export, and long copy-paste repair prompt are behind a one-time Stripe checkout. Referral attribution is preserved through the signed AISO referral cookie when present.
The detail rows, provider evidence, repair prompt and PDF export are behind a one-time unlock. The preview cards above stay visible.