bull-ai.engine
    Bull AI Labsbullailabs:~$
    Engineered AEO

    The engineering layer for AI Search visibility.

    $ bullai probe yoursite.com --agent-class
    [1/4]ChatGPT-User→200 OK · 0.4 kB
    ✗ FAILclient-side render - agent reads an empty shell
    [2/4]Claude-User→403 Forbidden
    ✗ FAILCloudflare default AI-bot rule - blocked at the edge
    [3/4]Perplexity-User→timeout 10.0s · TTFB 8.2s
    ✗ FAILfetch abandoned - slow origin, no citation
    [4/4]Googlebot→200 OK · rendered
    ✓ PASSfine on Google - invisible where buyers now ask
    $ bullai fix --sprint-01
    → ship SSR /pricing .........✓ ChatGPT-User re-probe: 200 OK · 41 kB

    Engineering, not more content. The queries that close deals are answered by a live fetch of your page at question time - and a fix ships to the next fetch, not the next training run.

    Not a dashboard you stare at. The engineering layer that moves the score.

    🔒app.bullai.tech
    LiveChatGPT3m ago"What CRM has the best AI features for sales teams in 2024?"High impact
    Overall AI Score
    7/100
    Visible Prompts
    0/45
    Losing to Competitors
    45/45
    Critical (<40)
    45/45
    5 prompts marked N/A. These prompts mention "Attio" by name — AI engines will always return the brand, so they're not a real visibility signal.
    Prompt Heatmap
    Each cell = 1 prompt. Color = average across 5 engines.
    ≥7055-6940-54<40
    1
    5
    2
    12
    3
    10
    4
    8
    5
    6
    6
    13
    7
    11
    8
    N/A
    9
    7
    10
    5
    11
    12
    12
    10
    13
    8
    14
    6
    15
    13
    16
    11
    17
    9
    18
    7
    19
    5
    20
    N/A
    21
    10
    22
    8
    23
    6
    24
    13
    25
    11
    26
    9
    27
    7
    28
    5
    29
    N/A
    30
    10
    31
    8
    32
    6
    33
    13
    34
    11
    35
    9
    36
    7
    37
    N/A
    38
    12
    39
    10
    40
    8
    41
    6
    42
    13
    43
    11
    44
    9
    45
    N/A
    46
    5
    47
    12
    48
    10
    49
    8
    50
    6
    "lightweight CRM for venture-backed startups"
    DISCOVERYscore 6

    Measured across 50 prompts and every major engine - not vibes.

    Bull AI Labs rebuilt how every AI engine sees Freeletics. Within a quarter we went from being absent in answer surfaces to being the cited choice for our category.
    Confidence Udegbue, VP of Product at Freeletics
    Confidence UdegbueVP of Product · Freeletics
    AI Search Visibility
    InvisibleCited
    Consistently over Nike, Peloton, Apple Fitness+ across 4 engines in one quarter
    Live · Q1 2025
    Two-Fetch Diff · live

    Ranking on Google doesn't mean ranking in AI.

    Googlebot renders your JavaScript. AI crawlers don't. Run any URL through the two-fetch diff and see exactly what ChatGPT, Claude, and Perplexity can — and can't — read on your site.

    Two-Fetch Diffidle
    raw fetch · no JS
    NOT VISIBLE
    $ fetch --no-js https://www.freeletics.com
    ChatGPTClaudeGeminiPerplexity
    AI crawlers can't read this
    Submit a URL to probe
    ↓ With JS executed (Googlebot)
    rendered · JS executed
    VISIBLE
    $ render --js https://example.com
    <h1>your-site.com</h1>
    <main>full page content, headings, schema…</main>
    GooglebotChatGPTClaudeGeminiPerplexity
    All engines read this
    Googlebot executes JS — AI crawlers don’t
    Raw fetch (what AI crawlers see) vs. rendered (what Googlebot sees) · probed via audit-render

    Not a tool. A stack for AI search visibility.

    Six product surfaces, one operating picture. Diagnose the crawl gap, score every engine, ship the fixes, measure the lift.

    Two-fetch diff

    See what AI crawlers actually see.

    Catch invisible content before it costs you citations.

    React SPA (unoptimized)
    EMPTY
    yourapp.com/pricing
    <div id="root"></div>
    Loading…
    3 words indexed
    3 words indexed
    Bull AI Labs engineered (SSR + AEO)
    INDEXED
    yourapp.com/pricing
    GPTBot · live fetch 200 OK
    ChatGPT answer
    "For engineered AEO, teams use yourapp.com to make pages readable by LLM crawlers."
    ↳ cited: yourapp.com/pricing
    1,284 words indexed
    Open

    Engine × Signal Matrix

    51 signals graded. 9 engines. One operating picture.

    Every cell is a citation opportunity, scored continuously.

    CGPTCLDPPLGEMAIOCOPGRKMETYOU
    S1
    S2
    S3
    S4
    S5
    S6
    S7
    CRIT HIGH MED N/A
    Open

    Sprint board

    Engineering tasks pushed as PRs.

    Bi-weekly cadence, scoped to revenue impact, reviewable as pull requests.

    TODO 2
    In Review 2
    Shipped 3
    PR-126
    Patch robots.txt → allowlist ClaudeBot, PerplexityBot
    PR-127
    Add llms.txt with capability declarations
    PR-123
    SSR pre-render /pricing and /features for bot access
    PR-128
    Ground entity to Wikidata Q-node via sameAs
    PR-129
    Migrate React SPA → Next.js App Router (SSR)
    PR-130
    Expose .well-known/mcp.json catalog endpoint
    PR-131
    Fix hydration mismatch blocking
    Open

    Citation map

    Track citation share across every engine.

    ChatGPT, Claude, Perplexity, Gemini, AI Overviews, Copilot. Per-prompt, per-engine, weekly.

    Citation Share / 12 WK+12% WK/WK
    58%51%44%32%
    70%35%0
    W01W06W12
    ChatGPT
    Perplexity
    Claude
    Gemini
    Open

    Frontier signals

    The signals that will define 2027.

    MCP-readiness, agentic commerce surface, .well-known endpoints. Tracked before competitors know they exist.

    Frontier Scan
    2027 Readiness
    3/5
    /.well-known/ai-plugin.jsonDETECTED
    /.well-known/mcp.json●MISSING
    /robots.txtDETECTED
    /agent-surface●MISSING
    /openapi.jsonDETECTED
    Frontier Readiness3 of 5 Detected
    Open
    51signals graded
    9engines tracked

    The platform

    From audit to citation in one bi-weekly cycle. Bull AI Labs is the engineering layer between your technical surface and every AI engine that decides what gets cited.

    01Your technical surface
    MCP EndpointMCP
    Agents.mdAGENTS.MD
    .well-known.WELL-KNOWN/
    robots.txtROBOTS.TXT
    RenderingSSR / RSC / SSG
    Bot AccessROBOTS / UA
    Entity GraphWIKIDATA / KG
    Content ArchCHUNKS / BLUF

    Bull AI Labs holds a point of view on every protocol surface that will determine AI citation — from chunk architecture to the MCP layer where agents will invoke instead of browse. 51 signals graded across the surfaces we measure today; MCP coverage is POV, not shipped probing.

    02Bull AI Labs engineers it
    BULL AIengineering layer
    01 / 03
    51 signals graded
    Diagnosed

    Across every surface above — rendering, entity, edge, community, reviews, agents.

    02 / 03
    Bi-weekly
    Engineering PRs

    Shipped as code, not slide decks. Two-week cycles, measured outcomes.

    03 / 03
    9 engines
    Measured

    Citation share tracked weekly across every major AI surface in the market.

    03AI engines cite you
    ChatGPT
    Claude
    Perplexity
    Gemini
    AI Overviews
    Copilot

    Your brand becomes the cited choice across every major engine — measured, not assumed.

    The Bull AI Labs methodology

    51 signals graded. 9 AI engines. Tracked continuously.

    Click any cell to see what the platform measures. Filter by engine to see the signals that move citation share for that pipeline specifically.

    THE ENGINE × SIGNAL MATRIX

    Every signal, every engine, at a glance.

    Each cell shows how that engine actually consumes this signal - match, downgrade, or N/A. Click any failing cell to jump to its detailed finding below.

    SIGNAL
    ChatGPTChatGPT
    ClaudeClaude
    PerplexityPerplexity
    GeminiAI Overviews
    GeminiGemini
    ChatGPTCopilot
    		IMPACT
    PER-ENGINE STATEBLOCKEDBLOCKEDBLOCKEDBLOCKEDBLOCKEDBLOCKED
    CSR/SPA shipping empty HTML to bots
    CSR_SPA_EMPTY_BODY
    Hydration > 2.5s (content invisible at crawl)
    HYDRATION_TIMING_OVER_2_5S
    Wikipedia page absent (ChatGPT corpus gap)
    WIKIPEDIA_ABSENT_CHATGPT
    Reddit presence absent (Perplexity citation gap)
    REDDIT_ABSENT_PERPLEXITY
    Wikidata Q-node absent (entity not resolvable)
    WIKIDATA_QNODE_ABSENT
    Organization schema missing
    ORGANIZATION_SCHEMA_MISSING
    sitemap.xml missing or invalid
    SITEMAP_MISSING_OR_INVALID
    Alt text missing in bulk (image cite gap)
    ALT_TEXT_MISSING_BULK
    BLUF answer missing above the fold
    BLUF_ANSWER_MISSING
    Comparison table absent (citation magnet)
    COMPARISON_TABLE_ABSENT
    CRITICAL HIGH MEDIUM DOWNGRADED N/A
    How the engines actually cite

    Different engine, different pipeline, different citation diet.

    • ChatGPT cites Wikipedia more than any other source
    • Perplexity and AI Overviews lean heavily on Reddit
    • AI Overviews surfaces YouTube prominently
    • Claude favors documentation, .gov, and academic sources

    Sources: Profound (2025), Ahrefs, Vercel/MERJ (2024). Updated quarterly.

    Most AEO playbooks apply one motion to every engine. That is why citation share does not move.

    08 / THE FRONTIER LAYER

    AI agents don’t browse.
    They invoke.

    The companies that get cited in 2027 aren’t the ones with the best content. They’re the ones whose products AI agents can actually use.
    Bull AI Labs audits your MCP surface across five dimensions — surface coverage, tool description quality, capability completeness, schema rigor, and error clarity — then ships the engineering work to fix every gap.
    Universal scoring. Industry-specific loops. Every vertical covered.

    bullai-probe · mcp-readiness
    $ bullai-probe --mcp-readiness https://acmestore.com
     
    ▸ MCP DISCOVERY
      ✓ /.well-known/mcp.json           FOUND   (manifest v2026-03-26)
      ✓ /api/mcp                         FOUND   (Shopify Storefront MCP)
      ✗ /api/mcp/customer                404     (Customer Account MCP)
      ✗ /api/mcp/checkout                404     (Checkout MCP preview)
      ✓ HTML meta declaration            FOUND
      ✓ robots.txt allowlist             5/7 AI crawlers permitted
     
    ▸ PROTOCOL SURFACE
      Transport:        HTTP + SSE (correct)
      Auth:             OAuth 2.1 with PKCE
      Spec version:     2026-03-26 (current)
      Capabilities:     resources, tools, prompts
     
    ▸ TOOL SURFACE (12 declared)
      ✓ search_products                  desc: well-formed, 142 tokens
      ✓ get_product_details              desc: well-formed, 98 tokens
      ⚠ add_to_cart                      desc too generic ("adds item")
      ⚠ get_inventory                    pagination schema missing
      ✗ checkout                         NOT EXPOSED — conversion blocker
      ✗ create_return                    NOT EXPOSED — retention break
      ✓ get_order_status                 well-formed
      ⚠ apply_discount                   errors not agent-readable
     
    ▸ CAPABILITY LOOP (ecommerce)
      ✓ browse         search_products, get_product_details
      ✓ cart           add_to_cart, get_cart
      ✓ auth           login, get_customer
      ✗ checkout       NO TOOLS — agent task fails at conversion
      ✗ post_purchase  NO TOOLS — no return, refund, or tracking surface
     
    ▸ AGENT DISCOVERABILITY
      ChatGPT crawl:           ✓ ingested 2026-05-14
      Claude Desktop:          ✓ MCP server registered
      Gemini Agentic Mode:     ⚠ partial (Customer MCP missing)
      Copilot Shopping:        ✗ not registered
     
    ▸ MCP READINESS SCORE: 58 / 100
      Surface coverage          61/100   ⚠
      Tool description quality  72/100   ⚠
      Capability completeness   34/100   ✗   (no checkout, no returns)
      Schema rigor              68/100   ⚠
      Error clarity             54/100   ⚠
     
    ▸ CRITICAL GAPS
      → Checkout MCP not deployed. Agents browse but cannot buy.
      → No returns surface. Post-purchase agentic loop terminates.
      → 4 tools have generic descriptions. Agents won’t dispatch.
     
    ▸ ENGINEERED SPRINT (4 PRs)
      PR-1   Deploy Shopify Checkout MCP preview          [critical]
      PR-2   Rewrite tool descriptions for agent dispatch [high]
      PR-3   Add Customer Account MCP                     [high]
      PR-4   Idempotency keys on cart manipulation tools  [medium]
     
      ESTIMATED CITATION LIFT:  +23pp ChatGPT, +31pp Perplexity Shopping
      ESTIMATED AGENTIC CONVERSION SURFACE:  $0 → addressable
    ECOMMERCE
    Browse → cart → checkout → returns
    SAAS
    Discover → query → act → integrate
    CONTENT
    Search → retrieve → recommend → subscribe
    DEV TOOLS
    Discover → diagnose → act → integrate
    PROFESSIONAL
    Research → draft → review → file
    10+
    MCP surfaces in scope

    /.well-known/, /api/mcp, Shopify storefront, agents.md, and more — every known agentic surface we track.

    5
    Optimization dimensions

    Surface coverage, tool descriptions, capability completeness, schema rigor, error clarity.

    WEEK-ZERO
    Coverage

    MCP, agents.md, .well-known — in scope before mainstream tooling exists.

    Customer proof

    Engineered AEO, measured in citations.

    Each result is a real customer outcome, scored against a fixed prompt set across nine AI engines.

    Freeletics · 12-week deployment

    Freeletics captured 68% AI search visibility in 12 weeks. Past Nike, Apple, and Peloton.

    68%
    AI visibility shareacross 135 fixed prompts
    +53 pts
    Lift from baselinemeasured, not modeled
    9 / 9
    Engines wonout of 9 tracked
    12 wks
    Time to outcomeplatform deployment
    ChatGPT response ranking Freeletics ahead of Nike, Apple, and Peloton for a fitness-app recommendation prompt.
    What ChatGPT returns when buyers ask for fitness app recommendations.
    9%47%+38 pts
    Citation share across 120 newsletter-discovery prompts

    10 weeks. Rendered-HTML fixes and Reddit/G2 entity reinforcement deployed through the platform. Measured on a fixed creator-tooling prompt set across all 9 engines.

    Beehiiv · Head of Growth
    11%52%+41 pts
    AI Search share across 180 restaurant-ordering prompts

    14 weeks. Schema rebuild, location-page rendering, and Perplexity-pipeline tuning deployed continuously. Share of voice across the restaurant-tech prompt set on 9 engines.

    Lunchbox · Director of Brand
    Confidence Udegbue, VP of Product at Freeletics
    Bull AI Labs rebuilt how every AI engine sees Freeletics. Within a quarter we went from being absent in answer surfaces to being the cited choice for our category.
    Confidence UdegbueVP of Product · Freeletics
    Reclaim your share of AI Search

    See the engineering gap between you and the brand AI is citing instead.

    No credit card · Audit in under 2 minutes · 9 engines scanned

    Found on AI — monthly underdog teardown.
    Subscribe →

    We use cookies

    We use cookies and similar technologies to enhance your browsing experience, analyze site traffic, and provide personalized content. By continuing to use our site, you consent to our use of cookies.

    •