Quickstart
Go from a domain to a useful combined SEO and AI visibility workspace.
A useful first scan depends on the inputs supplied before it starts. The canonical domain controls crawl scope, while the brand profile gives the GEO workflow enough context to recognize names, products, competitors, audiences, and markets. A focused prompt set is more valuable than a large collection of loosely related questions.
The initial setup is complete when the project has both website evidence and answer evidence that can be reviewed together. After the scan finishes, confirm coverage before interpreting scores, open individual issues and responses, and preserve the selected models, prompts, and market context when sharing the result.
For the first run, optimize for clarity rather than maximum volume. Use a small prompt set that represents important customer decisions, verify that the correct hostname is in scope, and select only the AI surfaces you intend to compare. This creates a readable baseline. Additional prompts, competitors, markets, and experiments can be added after the team understands how the initial evidence is organized.
Before you start
Have the canonical domain, the brand name, target market, main products or services, and a shortlist of competitors ready.
Decide who owns the project setup and who will review the first result. Collect access requirements, confirm the public hostname, and agree on the market and competitors before starting. Changes made after the baseline can alter the population, so record the initial configuration for later comparison.
- 1
Create your project
Add the canonical hostname. Sight normalizes the address and creates a project-scoped workspace.
Use the public canonical hostname without a path, tracking parameters, or alternate campaign domain. Confirm that redirects resolve to the intended site and that the project belongs to the correct organization before continuing.
- 2
Complete the brand profile
Add the description, industry, identity terms, products, personas, markets, language, country, and competitor context.
Write identity terms that are specific enough to recognize the brand without matching unrelated language. Include current products, audiences, and markets, then review competitor suggestions rather than accepting them as authoritative.
- 3
Prepare the prompt set
Create topics, review generated suggestions, activate the most decision-relevant prompts, and select available AI surfaces.
Start with a stable baseline that covers the decisions the audience actually makes. Remove duplicate wording, tag prompts by topic or intent, and document any location or language context that can change the answer.
- 4
Verify the site
Use the provided verification method to prove control of the domain. Verification unlocks scheduling and the public score badge.
Add the provided value using the exact supported method and confirm it is publicly reachable from the canonical hostname. CDN caching, redirects, or an incorrect environment can delay detection even when the value exists somewhere in the deployment.
- 5
Run the first scan
Sight crawls and renders the website, runs SEO checks, probes selected AI models, collects sources, scores the evidence, and builds the workspace.
Keep the initial configuration intentionally small enough to inspect. Watch the scan phases, confirm that expected pages and prompts were included, and record any degraded model or crawl behavior before using the result as a baseline.
Scans use plan limits
Page depth, rendered pages, performance checks, prompt count, model fanout, competitors, and AI budget vary by plan. See Limits and credits for the current code-backed values.
Review the effective plan before the first scan and again before comparing projects. A larger site or prompt set may be sampled differently, which changes what the resulting scores can represent.