# Your first LLM Citely project Canonical: https://llmcitely.com/docs/quickstart Publisher: LLM Citely Reviewed: 2026-10-31 Set up a brand, approve a small question baseline and learn how to read the resulting evidence. Confirm collection readiness before starting a run. Six platforms have passed local answer-and-source collection tests: ChatGPT, Gemini, Google AI Mode and Google AI Overviews for website answers; Perplexity and Claude for model API replies. Production rollout is pending. Region and search activity must be confirmed per sample before including it in comparable visibility metrics. [View platform readiness](https://llmcitely.com/platforms#platform-readiness). ![Start with a question you can review. The first useful outcome is a clear baseline and one supported next task.](https://llmcitely.com/images/editorial/first-project-walkthrough.svg) The first useful outcome is a clear baseline and one supported next task. Production platform acceptance is pending. Local examples are not measured customer results. Original explanatory diagram; illustrative counts and states are not customer measurements. ## Decide what this first project needs to answer Choose one client brand and the public domain it owns. Write down the market, language and the decision the project should support. A useful first task might be reviewing software-comparison questions for one buyer segment. Keep the first question list small enough that someone can inspect every answer. Separate unbranded discovery from questions that already name the brand. You can expand the set after reviewing the first evidence; changing it later requires a labelled baseline change. The [buyer-question guide](https://llmcitely.com/guides/choose-buyer-prompts) includes the selection criteria. The public [question builder](https://llmcitely.com/tools/geo-prompt-generator) produces editable drafts, not a measured database of customer demand. ## Create the project and confirm its inputs Open the workspace and create a project using the brand's public website. Review the proposed name, domain, brand aliases and competitors before accepting them. A suggested company is not automatically a verified competitor or a source you own. Next review the questions. Keep the exact wording, intended buyer task and approval reason. For each question, decide whether it is in scope now. Reject leading questions that only repeat your own positioning, and keep a note explaining the rejection. This public walkthrough describes the current local workflow. It does not create an account, grant provider access or claim a production collection has already passed acceptance. ## Check the surface and point allowance Before starting collection, open [platform readiness](https://llmcitely.com/platforms#platform-readiness). A configured adapter and a production-accepted surface are different states. Perplexity and Claude model API replies have passed local testing. They are labelled separately from website answers and must not be presented as monitoring their consumer websites. Review the workspace's available points and the run's quote. The confirmed ordinary collection rate is one shared point. A higher-cost operation needs its own supported quote; it must not be treated as an ordinary one-point run by assumption. Read [points and usage](https://llmcitely.com/docs/points-and-usage) before scheduling repeated work. The public pricing planner is an estimate of workload and package fit, not proof that a provider route is enabled. ## Open one answer before interpreting the summary When a collection completes, inspect its question, observed surface, time and readable answer. Check the context of a brand mention and the destination of an explicit source link. A provider name in the text does not automatically mean the brand was recommended. If the request fails or is still collecting, keep that state visible. It is not a valid answer that omitted the brand. The [status guide](https://llmcitely.com/docs/collection-status-and-retries) explains how to investigate these states. Use the [measurement walkthrough](https://llmcitely.com/guides/how-to-measure-ai-visibility) to understand coverage and denominators before sharing a percentage. ## Finish with a small, checkable handoff Pick one finding that the original answer and source can support. Record the evidence, assign an owner and write an acceptance condition in the [action plan](https://llmcitely.com/templates/geo-action-plan). If the work is editorial, prepare a [content brief](https://llmcitely.com/templates/ai-search-content-brief). If it is technical, follow the [audit walkthrough](https://llmcitely.com/guides/run-a-geo-audit). Recheck the actual changed page before interpreting later AI observations. Your first useful outcome can be a verified question baseline and one well-supported task. A registration, started run or attractive example chart is not evidence of an improvement in search visibility. ## Questions, answered ### Can I follow the first steps without live provider access? You can review public tools, example inputs and a proposed question baseline. A real collection still needs a configured, accepted surface and available allowance. ### How many questions should the first baseline contain? Use a set small enough to review individually. The right number depends on the buyer tasks and evidence budget; a larger count is not automatically a better sample. ## Sources and further reading - [Analyze AI: an example of task-oriented onboarding documentation](https://www.tryanalyze.ai/docs/getting-started/quickstart) Examples marked illustrative are not customer measurements. ## Continue with the next step - [Understand the collection states](https://llmcitely.com/docs/collection-status-and-retries) - [Check shared points and quotes](https://llmcitely.com/docs/points-and-usage) - [Review buyer-question selection](https://llmcitely.com/guides/choose-buyer-prompts) - [Prepare the first action](https://llmcitely.com/templates/geo-action-plan) ## Your next step [Understand the collection states](https://llmcitely.com/docs/collection-status-and-retries)