Frequently Asked Questions

Is this just prompting ChatGPT?

No. Here's the exact difference:

• • Prompting means typing a question into ChatGPT's interface and getting one answer back. That answer changes every time you ask, and you can't measure it statistically.
• • Probing (what we do) means calling the assistant-search APIs directly — OpenAI's web search, Perplexity Sonar, Gemini grounded — and recording the exact sources, citations, and excerpts they return. We run the same query multiple times to confirm consistency, and we label every sample with the model ID, timestamp, and source list.

The difference matters because prompting gives you vibes ("I think ChatGPT likes this page"); probing gives you data ("ChatGPT cited this page in 3 of 5 samples for this query, and here are the 5 raw responses").

What's a "single sample per lane"?

Every free scan and deep audit is a dated snapshot, not ongoing monitoring. For each intent, we probe each of the 3 lanes (ChatGPT, Perplexity, Gemini) once and record the exact answer, sources, and citations. We don't run the query 100 times and average it — we run it once, capture the full response, and show you exactly what we got.

This is deliberate: we're showing you what assistants said on this date for this query, not claiming it's the only answer they'll ever give. The methodology section of every report states this limitation plainly.

Does the deep audit guarantee I'll rank #1?

No. We measure citation influence — whether assistants cite you and where your citations come from — not placement guarantees. The deep audit shows you which pages assistants read to answer your category, which you own vs don't, and what you can build or place to contest the gaps. Whether that moves your citations is something we measure in the 30-day re-scan, not something we guarantee.

What can't you control?

A lot. Here's the honest list:

• • We can't control which pages OpenAI, Perplexity, or Google choose to crawl, index, or cite.
• • We can't control how often they re-crawl your pages or how long it takes for a change to show up in answers.
• • We can't force assistants to cite you — we can only influence it by optimizing your pages and placement.
• • We can't guarantee backlinks from third-party sites will land (Autopilot can propose and draft them, but publishing is up to the site owner).
• • We can't control your competitors — if they're optimizing faster than you, they'll stay ahead.

What we can control: the Source Placement Map (finding every page assistants cite), the Fix Kits (writing the capsules and schema you need), and the execution loop (Autopilot publishing and iterating on what moves). Everything else is influence, not control.

What does the $79 deep audit actually buy me?

You get the execution plan that Autopilot will act on:

• ✓ 25 intents — 10 you supply, 15 we expand from your site and category
• ✓ Full Source Placement Map — every page assistants cited (usually 40+), classified by ownership (yours, competitor-owned, third-party placement targets)
• ✓ Competitor teardowns — the top 1-2 pages beating you, fetched and analyzed for shape gaps (what they have that you don't)
• ✓ Paste-ready Fix Kits — answer capsules + JSON-LD schema blocks you can deploy immediately
• ✓ Leverage ranking — fixes prioritized by impact (intents unlocked per fix)
• ✓ 30-day re-scan — we re-run this exact intent set in 30 days and email you the before/after delta

See a real one: sample deep audit for CostBench.com

Can I get a refund?

Yes. Deep audits have a 7-day no-questions refund window from purchase. Email [email protected] with your order email and we'll process it same day. You don't need to explain why — "I want a refund" is enough.

Why is the free scan capped at 5 intents?

Cost. Each intent × 3 lanes = 3 API calls (OpenAI, Perplexity, Gemini). At 5 intents that's 15 calls, ~7¢ in credits. We can afford to give that away. At 25 intents (the deep audit tier) it's ~$1.40 — not sustainable at $0. The free scan shows you enough to decide if the deep audit is worth $79.

How much will Autopilot cost?

We don't know yet — it's not launched. Pricing will likely be usage-based (per intervention approved and executed) rather than a flat monthly fee. Design partners will get founding pricing spelled out in their agreement before anything runs. We'll update /pricing when we know more.

What data do you collect?

For free scans and deep audits (no account):

• • Your domain and the intents you submit (we need these to run the scan)
• • Your email (to send you the report link)
• • The raw API responses from OpenAI, Perplexity, and Gemini (this is your report data)
• • Page fetches of your public pages and competitor pages (for shape analysis in deep audits)

We don't collect passwords, API keys, CMS credentials, or backend access. We only read public pages.

Do you sell my data?

No. We don't sell domain lists, intent lists, emails, or report data. We use your data to generate your report, send it to you, and (if you paid for a deep audit) to send the 30-day re-scan. That's it.

Can I delete my data?

Yes. Email [email protected] with the email you used to request the scan/audit, and we'll delete your report and associated data within 7 days. Note: if you paid for a deep audit and delete before the 30-day re-scan, we can't send you the re-scan (we won't have your original baseline to compare against).

What will Autopilot need access to?

When Autopilot launches, it will need:

• • CMS API access (write) to publish answer capsules and schema to your pages
• • Read access to your existing pages (to detect what's already there and avoid duplication)
• • No backend/database access — Autopilot only touches content via your CMS API, same as a human editor would

Every connection and every publish action will require your explicit approval. Read the full details in the operating contract.

What if Autopilot hallucinates?

Good question. Here's the honest answer:

LLMs do hallucinate on platform-facts shapes (our own evals show ~33% fabrication rate on "how does X work" content). So Autopilot is built around deterministic grounding wherever possible:

• → Answer capsules quote only your existing copy (we fetch your page, extract claims, and rewrite for question-first format — no fabrication)
• → Schema blocks derive from actual page parses (Product schema = your pricing table data; FAQ schema = your existing FAQ text)
• → llms.txt generation uses real crawl logs (we don't guess what you want crawled — we show what assistants actually requested)
• → Every intervention shows you the diff before publishing (you see exactly what changes)

Where we can't ground deterministically (e.g., drafting a backlink pitch email), Autopilot flags it as "needs review" and shows you the draft before sending. The rule: if it can't be verified against your own content or real crawl data, it doesn't auto-publish.

Full mechanics in the operating contract.

Can Autopilot break my site?

Autopilot can only do what your CMS API allows. If your CMS has rollback/versioning (most do), you can undo any change Autopilot makes. If it doesn't, Autopilot will show you a backup of the original content before publishing the new version.

Autopilot cannot delete pages, drop databases, change DNS, or access your backend — it only writes content via your CMS API, same as a human editor logging into your CMS.

When will Autopilot launch?

We're building it now with design partners. No public date yet. If you want early access and founding pricing, apply as a design partner. We'll email you when we're ready to onboard the next cohort.

Can I turn Autopilot off?

Yes. You can pause or cancel at any time. When paused, no new interventions run — but you keep the Source Map, Fix Kits, and re-scan data from your deep audit. When you cancel, we revoke CMS API access within 24 hours.