Run your first AI Scan

An AI Scan asks dozens of prompts to multiple AI models and measures how your brand appears.

Last updated: 2026-05-05

TL;DR — An AI Scan asks each of your selected prompts to each of your selected AI models, then measures how (and whether) your brand appears in the responses. Run one from the Dashboard → Run AI Scan button. Cost = (prompts × models) credits, billed at $1 = 1 credit. Google AI Overview is selected by default.

What an AI Scan actually does

When you launch an AI Scan, Pranas executes a controlled experiment:

• Inputs: a set of prompts (from your prompt library) and a set of AI models.
• Process: for every (prompt × model) pair, Pranas sends the prompt to that model fresh, with no prior context, exactly as a real user would on ChatGPT, Perplexity, Google, Claude, or Gemini.
• Outputs: the model's full natural-language response, which Pranas parses to extract:
  • Whether your brand was mentioned
  • Where it ranked relative to other brands
  • Which competitors appeared
  • Which sources/URLs the model cited
  • Sentiment and context around your brand
  • Topic and category positioning

This is the foundational measurement that powers every dashboard metric: AI Score, Share of Voice, Top Brands, Co-occurrence, Topic Visibility, Source Influence, and Brand Perception.

Why it matters

You can't optimize what you can't measure. Generative engines like ChatGPT and Google AI Overviews are increasingly the first surface prospects use to discover, compare, and qualify products. If your brand is missing — or worse, mentioned in a misleading or outdated context — you're losing pipeline before a human ever visits your site. AI Scans give you the empirical, repeatable, time-series data you need to know exactly where you stand and what to fix.

Step-by-step: Run your first scan

1. Confirm prerequisites:
   • At least one brand exists.
   • At least one prompt is in your prompt library — if not, generate prompts first (see Generate prompts).
   • You have credits available.
2. Open the Dashboard and click Run AI Scan in the topbar (or Results → New Scan).
3. The Scan dialog opens.
4. Select prompts — for a first run, pick 10–25 prompts spanning Unaided, Aided, and Comparison types. More prompts = more statistical confidence, but more credits.
5. Select AI models — Google AI Overview (Gemini 3) is pre-selected as the default. Add ChatGPT, Perplexity, Claude, or others depending on your plan tier. Premium models may be locked on lower plans.
6. Review the credit estimate shown at the bottom of the dialog.
7. Click Run Scan.

What happens after you click Run

Your scan goes into the simulation queue and runs in the background. You can keep using Pranas, navigate away, even close the browser — execution continues server-side. You'll see live progress in the topbar and on the Results page. When it completes:

• A First Scan Celebration banner appears (for your very first scan).
• The dashboard widgets refresh with new data.
• A summary card appears with key metrics.
• Detailed per-query results are accessible in Results.

For larger scans, the queue processes in batches and may take from 1 minute (small scan, fast model) to 10–20 minutes (large scan, slow premium model).

Worked example

Maya runs her first scan for Acme Feedback:

• Prompts: 20 (12 unaided like "best customer feedback tools for B2B SaaS", 5 aided like "is Acme Feedback good for product teams", 3 comparison like "Canny vs ProductBoard vs Acme Feedback")
• Models: 3 — Google AI Overview, ChatGPT (GPT-5 mini), Perplexity
• Estimated cost: 60 executions × per-execution cost = ~$X credits

She clicks Run Scan. Within 4 minutes the scan completes. The dashboard now shows an AI Score of 47 and a Share of Voice of 12% — Acme appears in 18 of 60 responses, usually ranked 3rd–5th behind Canny and ProductBoard. She knows exactly where she stands.

Credits and cost transparency

Pranas uses a strict $1 = 1 credit ratio. Each (prompt × model) execution consumes credits proportional to the model tier and the length of the prompt and response.

• Estimate is shown before launch — you'll never accidentally burn credits.
• Premium models cost more credits per execution (e.g. GPT-5, Gemini 3 Pro) than fast/cheap models (e.g. GPT-5 nano, Gemini Flash Lite).
• Failed queries are auto-retried once — retries are not double-billed.
• No grace period: if your balance runs out mid-scan, remaining queries fail and you'll be prompted to top up.

See Credits per scan for the full pricing breakdown.

Edge cases & gotchas

• No prompts in library: the Run AI Scan button is disabled until you generate or import at least one prompt. Pranas will redirect you to Prompts with a guided message.
• Email not verified: scan execution is blocked until email is verified. See Create your account.
• Premium model locked: if you select a premium model your plan doesn't include, the dialog will show a lock icon and an upgrade CTA.
• Scan stuck in "Processing" for >30 min: rare, usually a model outage. See Scan statuses explained.
• Concurrent scan limit: each user has a per-user concurrent execution cap to prevent edge-function timeouts. If you're at the cap, new prompts wait their turn in the queue.

Troubleshooting

Related articles

• What is an AI Scan?
• Choose models for a scan
• Scan queue and retries
• Scan statuses explained
• Credits per scan
• Reading scan results
• Generate prompts

Still need help?

Email contact@pranas.co with your scan ID (visible on the Results page) and we'll investigate.