One JavaScript pixel plus a server-side NPM adapter ship FAQ schema, llms.txt, robots updates, and freshness signals to your site. Six sub-features keep it honest: deploy-approval workflow, automated verifier, auto-rollback on regression, AI bot-crawl tracker, and a cross-platform deployment dashboard. No developer queue. No surprise overwrites. No silent failures.
Every AI visibility tool on the market tells you what to fix. None of them fix it. You get a report that says “add FAQ schema to your top 20 pages,” “create an llms.txt file,” “update your Organization schema.” Then the report sits in a Google Drive folder for months while the developer queue stays 8 weeks out.
Autopilot closes that gap with six locked sub-features. F-1 detects AI bot crawls server-side via the NPM adapter — catching bots the JS pixel can't because it never renders. F-2 ships fixes through the hybrid pixel + NPM model. F-3 routes every fix through a deploy-approval workflow with tier-gated quotas and a 200-entry JSONB audit log. F-4 verifies every deployed fix every 10 minutes against your rendered HTML and auto-rolls back regressions. F-5 tracks which AI bots crawl which pages. F-6 unifies all of it into a cross-platform dashboard for agencies managing mixed installs across a client roster.
The result: your fixes go live the same day you approve them, and the system tells you the moment any one of them breaks.
Server-side telemetry via @topslot/server-pixel. Detects AI bot user-agents (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) at the request layer and emits npm_bot_crawl_signals. Catches bots that the JS pixel can't because it never renders.
Client-side JS pixel injects FAQ schema, llms.txt link, robots.txt update, and dateModified freshness signals at render time. Server-side adapter unifies signals across both layers. Adapters ship for Next.js, Express, Remix, and Nuxt.
Every fix goes through pending → approved → deployed (with reject and rolled_back as terminal states). Tier-gated quotas: Starter 1/day, Growth 1/day, Agency 5/day. JSONB audit log of every state change, capped at 200 entries. You always know what was deployed and by whom.
Every 10 minutes, the verifier checks every deployed fix against your rendered HTML. With server-side adapter installed, missing fixes trigger auto-rollback + email alert. JS-only installs surface verified_with_warning instead. Alert throttled 24h per brand.
Dashboard at /dashboard/ai-bot-crawls. by_bot, by_day, and top_pages breakdowns show which AI models are crawling which pages. Detection-source view splits pixel / npm / static so you know how each crawl was caught.
Agency-tier unified view per brand. Three columns: JS pixel (configured + health + 30-day stats), NPM adapter, static fallback. 30-day stacked-bar timeline of detection sources. Designed for managing mixed-install configurations across a client roster.
Generates and injects FAQ structured data based on your content. FAQPage schema increases AI citation rates because it gives models pre-formatted question-answer pairs to extract.
Deploys full Organization structured data including name, logo, social profiles, founding date, and description. Helps AI models build accurate entity profiles of your brand.
Creates and maintains an llms.txt file at your domain root telling AI crawlers what your company does, what pages matter, and how to cite you. Updated automatically as your site changes.
AI models give strong preference to content updated within the last 90 days. The pixel injects dateModified schema and last-updated metadata so AI models treat your pages as current, not stale.
Configures robots.txt to allow AI crawlers selectively — GPTBot, ClaudeBot, PerplexityBot, Google-Extended — while keeping other bot policies intact. Per-page controls for sensitive sections.
Every change Autopilot makes is versioned and logged. F-3 audit log shows what was deployed, when it went live, and what it replaced. F-4 auto-rollback handles regressions; manual rollback is one click.
You know you need schema markup and llms.txt but you don't have a developer on staff. Autopilot handles the technical work; the deploy-approval workflow keeps you in control of every change.
Your developer queue is 8 weeks long. Marketing can't wait when AI visibility shifts weekly. Deploy fixes through the approval workflow without filing a Jira ticket; the verifier catches regressions automatically.
Install one pixel per client site and manage all deployments through F-6's cross-platform dashboard. White-label reports show clients exactly what was fixed, what was rolled back, and which AI models are actively crawling them.
Your engineers are building product, not updating schema. Autopilot keeps your marketing site optimized for AI search. Server-side adapters ship for Next.js, Express, Remix, and Nuxt — install in one PR, then never touch it again.
Pixel install works on every tier including Free.
Auto-fix deploy is available on Starter and above. Starter caps at 1 deploy per day; Growth at 1/day; Agency at 5/day.
F-4 auto-rollback on regression: Growth and Agency.
F-6 cross-platform deployment dashboard: Agency only.
▸ FOUNDER-LED ONBOARDING INCLUDED ON EVERY PAID PLAN
No. The pixel loads asynchronously after your page content renders. It adds less than 2 KB to your page weight and does not block the critical rendering path. Lighthouse scores remain unchanged.
F-3. Every fix goes through a state machine: pending → approved → deployed, with reject and rolled_back as terminal states. Quotas are tier-gated: Starter 1/day, Growth 1/day, Agency 5/day. Every state change is captured in a JSONB audit log capped at 200 entries per fix. You always know what was deployed, when, and by whom.
F-4. A verifier cron runs every 10 minutes against every deployed fix. If your install includes the server-side adapter and the fix is missing from the rendered HTML, the system declares regression_detected, auto-rolls back to the previous state, and sends an email alert. If you only have the JS pixel installed, the verifier surfaces verified_with_warning instead — degraded but not failed. Alerts are throttled to one per brand per 24 hours so you don't get an alert storm during a transient outage.
F-5. GPTBot, ChatGPT-User, ClaudeBot, Claude-Web, Google-Extended, PerplexityBot, Anthropic-AI, Applebot-Extended, FacebookBot, and more. Detection uses user-agent strings, IP-range checks, and behavioral patterns. The dashboard at /dashboard/ai-bot-crawls shows by_bot, by_day, and top_pages breakdowns so you know which AI models are actively crawling which of your pages.
F-6. A unified three-column view per brand: JS pixel (configured + health + 30-day stats), NPM server adapter, and static fallback. Includes a 30-day stacked-bar timeline showing detection-source breakdown across pixel / npm / static. Designed for agencies managing fixes across multiple client sites with mixed install configurations. Available on Agency.
Yes. Remove the script tag from your site header and all deployed fixes are removed on the next page load. Autopilot does not modify your source code or CMS database. Server-side adapter installs are removed by uninstalling the NPM package.
No. Autopilot detects existing schema markup on each page before deploying. If Yoast or RankMath already provides FAQ schema, the engine skips that page and focuses on pages without coverage. The deploy-approval workflow shows you the conflict explicitly so you can choose to keep, replace, or merge.
Run a free Scorecard and get a list of fixes Autopilot can deploy through the approval workflow. The verifier checks every one of them every 10 minutes after deploy.
Check your visibility, free →