Answers, in detail.

You request a demo via the button in the nav. We schedule a 30-minute call within a business day. On the call we walk through the platform on a real client site (we'll use one of yours if you give us permission), and at the end we activate a 14-day Pro trial on your account. No credit card required to start the trial.

Self-hosted (the Free tier) means you run the docker-compose stack on your own VPS, laptop, or VPC. You manage the Postgres, the Supabase, and the LLM API costs directly. Cloud (Pro/Agency/Enterprise) means we run that infrastructure for you — you just sign in. All four tiers have the same feature set; the difference is who runs the boxes and which LLM bills land where.

A list of tracked keywords, a list of competitor domains (optional), and read-access to your Google Search Console + Bing Webmaster Tools (we use the free APIs). Reviews come in automatically once you connect Google Business Profile / Yelp / Meta. Most operators are fully migrated within a week, with the AI Pulse + old-tool data running in parallel for the first month so you can validate.

First baseline audit + LLM suggestions are produced in under 5 minutes from clicking 'Run.' First weekly digest fires after 7 days. Meaningful keyword movement from acting on the suggestions typically shows up at 4–8 weeks (this is just SEO timing — no tool can shortcut Google's update cadence).

Cloud tier: no. You sign in, click through the onboarding wizard, and you're done in 15 minutes. Self-hosted tier: yes — you need someone comfortable with docker-compose and a bit of nginx. We have a step-by-step install guide and a Discord channel where the founder answers questions personally.

Yes. The /admin/seo/keywords page has a 'Bulk import' tab that accepts CSV with columns: keyword, target_url, intent, pillar. Or just paste your existing SEMrush / Ahrefs export — we'll auto-map the columns.

Three providers out of the box: OpenAI (gpt-4o-mini default, gpt-4o for premium tasks), Anthropic Claude (sonnet-4-5 / opus-4-5 / haiku-4-5), and local Ollama (any model you can pull — phi3:mini, qwen2.5:7b, llama3.x, mistral, etc.). Per-site provider choice means a client's strategist can run on Claude while a smaller client uses gpt-4o-mini.

Only the prompts and completions for the specific operations you trigger (suggestion generation, article drafting, etc.). We never send raw analytics, raw review text outside of what's needed for sentiment classification, or any field marked 'is_secret' in seo_settings. If you can't send any data to a third-party LLM, use the local Ollama provider — it runs on your machine and never makes external calls.

Yes. Configure PULSE_LLM_DEFAULT_PROVIDER=ollama and pull the model of your choice with `ollama pull qwen2.5:14b`. All LLM calls stay on your hardware. Performance trades off (CPU inference is 5-15× slower than gpt-4o-mini), but for compliance-heavy industries it's the right choice.

Analytics is cookie-less by design — no consent banner needed. Visitor fingerprint is a SHA-256 hash of UA + screen + timezone + language (no PII). DNT and GPC headers are honored — opt-out users are skipped entirely. Reviews + brand context are stored in your Postgres (your own DB on self-host, our managed DB on cloud). Right-to-delete: hit /admin/site-config and click 'Export + purge.'

You do. AI Pulse never claims ownership over generated articles, suggestions, or replies. The LLM provider's terms apply (OpenAI and Anthropic both grant you full ownership of outputs from their APIs).

Every LLM-generated suggestion is reviewable before it lands anywhere user-facing. Articles go to /admin/content as drafts (status='draft' — not published). Review reply drafts queue in /admin/profiles/reviews — you click Approve before they post. SEO suggestions appear in /admin/seo/suggestions with their reasoning visible — nothing auto-applies. The LLM is a drafting tool, not an autopilot.

14 days of Pro tier, activated at the end of the demo call. No credit card required. After day 14 the account auto-downgrades to Free tier (which keeps all your data — you just lose the cloud-hosting and can self-host) unless you choose to subscribe.

Yes — pay annually, save 2 months. Pro $490/yr (vs $588), Agency $1,990/yr (vs $2,388). Enterprise is always annual or longer with custom pricing.

Both. Each tier has a baseline number of sites + seats included; extra sites are $15/mo each, extra seats are $10/mo each. We weight pricing toward sites because that's what scales operational cost — adding a 10th site to your portfolio is the lever, not adding a 4th teammate.

On Agency tier, you can replace the AI Pulse logo + brand color with your own across the per-client portals. Clients see 'Your Agency' in the nav, your color, your subdomain (we provide a CNAME setup). Backend is still us; surface is you.

Cancel anytime, prorated refund for the unused portion of the month or year. No retention call, no exit interview, no friction. We'd rather hear honest feedback in an email than trap unhappy customers.

Yes. The Content engine has a one-click 'Publish to WordPress' button on every article draft. Auth via Application Passwords (more secure than user/pass). Drafts post to your WordPress as drafts (you click publish in WP). Future: Webflow, Ghost, MDX-on-GitHub.

GSC: yes — connect via service account JSON or OAuth (we have docs for both). Pulls daily ranking + impression + click data into the keywords dashboard. GA4: not yet (Pulse uses its own first-party analytics). On the roadmap for Q3 2026 if customer demand stays strong.

Yes — set POSTHOG_KEY in .env.local and pageview/event data also flows to PostHog in addition to AI Pulse's built-in analytics. Useful when your team already lives in PostHog for product analytics and you don't want to migrate.

n8n is first-class — we have a built-in publisher that signs payloads with HMAC. Zapier and Make: any AI Pulse event can trigger a generic webhook (see /admin/site-config/webhooks). The route is the same shape n8n's webhook node expects, so cross-platform routing works.

Yes. Connect a Slack incoming webhook URL in /admin/site-config/integrations. Configure per-rule routing (rank-regression alerts to #seo, negative-review alerts to #ops). Quiet hours per channel.

Content engine drafts in any language gpt-4o-mini / claude-sonnet supports natively (~30 languages with high quality). Reputation module classifies reviews in any language. Admin UI is English-only for v1; i18n on the roadmap.

AI Pulse generates 6-10 buyer-intent prompts per pillar (e.g. 'best Indian restaurants in Aurora IL for catering'). Each prompt is sent on a schedule to ChatGPT, Perplexity, and Google's AI Overview. We parse the cited sources from each response. If your site is cited, +1 to your visibility score for that prompt. Trends over time appear in /admin/citations.

Yes — three layers: (1) the auto-extracted brand context (industry / audience / tone) which AI Pulse infers from crawling your site, (2) per-site brand voice settings in /admin/seo/settings (custom tone, banned phrases, must-mention items, signoff), (3) per-suggestion-type prompt overrides for advanced users.

Cloud tier: managed Postgres (Supabase) in US-East by default; EU available on Agency+ for an extra $50/mo. Self-host: wherever you put it. Enterprise: custom data residency (US / EU / APAC) negotiable.

In progress — Type II report scheduled for Q3 2026. Until then we can share our security questionnaire responses + the auditor's NDA-friendly progress letter. Email hello@thoughtwavesoft.com to request.

Yes. Postgres on the cloud tier uses AES-256 disk encryption (Supabase default). All credentials stored in seo_settings (ad accounts, business profile tokens, etc.) are encrypted with PULSE_ENCRYPTION_KEY (AES-256-GCM) before the row is written. The key is stored separately from the database.

Available on Enterprise tier. Okta, Azure AD, and Google Workspace via OIDC / SAML. Provisioning via SCIM on request.

Yes. /admin/site-config has a 'Full export' button that produces a Postgres dump + asset bundle (zipped JSON of every article, suggestion, review, and analytics row). Re-import into the self-host docker-compose and you're independent of us. We'd hate it, but we'd rather you stay because you want to than because you're stuck.