Your customers are asking ChatGPT.
Make sure it answers with you.

BrandAnswers is a free, open-source kit that helps your brand get retrieved, cited and recommended in AI answers from ChatGPT, Google AI Overviews and AI Mode, Perplexity, Claude and Copilot. It runs inside the AI agent you already use, such as Claude Code, Cursor or Codex. It audits how AI crawlers see your site, then fixes what it finds in your code.

The kit has three parts. An Agent Skill holds the evidence-graded rulebook, reference guides and four zero-dependency scripts. An MCP server serves the live rulebook plus audit, answer-checking, JSON-LD and FAQ search tools, hosted at https://kit.brandanswers.ai/api/mcp or run yourself. And this site, with a living FAQ you can search in your browser. There is no account, no dashboard and no subscription.

Asking an LLM for FAQs gives you text, but BrandAnswers starts by checking whether AI crawlers can read your page at all. Many sites built with vibe-coding tools are client-rendered, so ChatGPT, Claude and Perplexity see an almost blank page. BrandAnswers fixes that first, then writes answers grounded in verified facts using evidence-graded rules.

It also covers what a prompt cannot: robots.txt and sitemap fixes, the commercial question shapes engines search (best X for Y, X vs Y, alternatives, pricing, is it legit), an answer checker, off-site research, since most recommendations come from third-party pages, and measurement by mention rate over repeated runs. And it says plainly what does not work, such as llms.txt.

Yes. BrandAnswers is free and open source under the MIT licence, with no account, dashboard or subscription. The skill and the hosted MCP cost nothing to use. The only optional cost is automated measurement, which uses your own Perplexity, OpenAI or Anthropic API keys and is billed by those providers at a small cost per run.

Your AI agent's own costs still apply, such as your Claude Code, Cursor or ChatGPT plan. If you prefer zero spend on measurement, paste your prompts into the AI apps by hand and let your agent tally the results. There is nothing to cancel and no trial period, because there is no subscription in the first place.

In Claude Code, run /plugin marketplace add jonwilliams88/brandanswers-kit, then /plugin install brandanswers@brandanswers. In other MCP clients, add the hosted server at https://kit.brandanswers.ai/api/mcp. In Cursor, Codex, Gemini CLI or VS Code Copilot, you can also add the skill to the client's skills folder.

For the skill, download /brandanswers-skill.zip from this site or run npx skills add jonwilliams88/brandanswers-kit, then place it in your client's skills folder. To add the hosted MCP in Claude Code: claude mcp add --transport http brandanswers https://kit.brandanswers.ai/api/mcp. Then ask your agent something like "audit my site for AI search".

Yes, in most cases. If your builder supports custom MCP servers, paste https://kit.brandanswers.ai/api/mcp into its settings. If not, ask an agent that has the MCP to audit your public URL, or run node scripts/audit.mjs <url> from the skill. Then apply the fixes in your builder, starting with rendering.

This matters because client-rendered apps, including older Lovable projects, Bolt and some v0 exports, can look blank to ChatGPT, Claude and Perplexity. If your builder cannot server-render or prerender, ask it to, or export the code to a repo where a coding agent can make the change.

Yes, in your agent, with your approval. Your agent makes the changes in your repo, such as fixing rendering, robots.txt, the sitemap and answer pages, and shows you the diffs. Your agent's normal permissions apply, so nothing is changed that you have not allowed. BrandAnswers itself never touches your code outside your agent.

Typical changes include switching key pages to server-rendering or prerendering, updating robots.txt to allow the AI search and user-fetch bots, adding a sitemap with lastmod, adding canonical URLs, writing answer sections into your own components as visible HTML, and adding Organization JSON-LD. Your agent re-runs the audit to prove each fix.

Yes. The skill runs locally in your agent, and the hosted MCP fetches only the public URLs you ask it to audit, stores nothing and has no accounts. The living FAQ on this site is searched in your browser, so your question is not sent anywhere. Your code never leaves your machine because of BrandAnswers.

The usual privacy terms of your AI agent and its model provider still apply to whatever that agent reads. If you run measurement, your prompts go to the AI providers whose keys you use, under their terms. The hosted audit also refuses private and internal addresses, so it cannot be used to probe your network.

It cannot guarantee that, and no tool honestly can guarantee citations or rankings. BrandAnswers removes the blockers that stop AI engines reading your site and improves your odds with answer pages and an off-site plan. Whether ChatGPT recommends you depends largely on what other sites say about you, which takes time.

Each fix is tied to a rule and an evidence grade, so you can see how strong the case is. Measure progress as mention rate over repeated runs, because AI answers change from run to run. Expect crawlability fixes to matter quickly for retrieval, and off-site work to matter most over months.

The search box needs JavaScript. Browse all 414 answers instead.