AI search optimization (GEO): get cited by ChatGPT and Google AI

What AI search optimization includes#

We shape the engagement around the actual brief. The proposal identifies which of these deliverables are included, who supplies the inputs and how each one is accepted.

• AI visibility audit across Google, ChatGPT, Perplexity, Claude and Copilot
• Crawler access fixes in robots.txt, CDN and firewall rules
• Answer pages for the questions buyers ask before contacting you
• Entity cleanup: one name, description and set of profiles across the web
• Map of the outside sources assistants cite in your category
• Agent-ready layer where it fits: markdown copies, llms.txt and MCP tools
• Monthly report with a fixed panel of buyer questions in English and Spanish

When to invest in AI search visibility#

Choose this when your website already works and you need to know how AI assistants describe your business, or why they cite a competitor instead. If pages are not indexed or the offer is unclear, start with our SEO service, since AI answers rest on the same foundations. If you want AI to do work inside your company, see AI automation.

How an AI search engagement runs#

We start with a dated baseline: crawler access per platform, indexing in Google and Bing, current Search Console and Bing numbers, and a fixed panel of buyer questions in English and Spanish. The audit ends in a short list of changes ranked by effect and effort, each with an owner and a check.

Then we implement what you approve: access rules, page structure, answer pages, structured data that matches the visible text, profile cleanup and, where it fits, the agent-ready layer. Each month we rerun the same panel and report what moved and what we would change next. Answers vary by platform, location and day, so we read trends, not single screenshots.

What gets a business cited in AI answers?#

Being reachable, being indexed and saying something worth repeating. Google states there are no additional requirements to appear in AI Overviews or AI Mode: the page must be indexed and eligible for a snippet, and no AI text files or special markup are needed. OpenAI says sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers, and Anthropic says blocking Claude-SearchBot may reduce visibility in Claude's search results.

So access comes first: a CDN or bot-protection rule can block a search crawler that robots.txt allows. Then content: Google's guidance for its AI features favors non-commodity content built on first-hand experience, not summaries of what is already online. Then evidence elsewhere: directories, reviews, press and partners that describe you accurately. Google warns that chasing inauthentic mentions does not help.

Nobody can guarantee a citation; the platforms decide what to show. Dardo sets every lever under your control and measures the effect the same way each month.

What does an AI visibility audit cover?#

Whether each assistant can reach your site, whether it understands who you are, and what it says when a buyer asks about your category:

Two problems are easy to miss: a security rule that blocks a crawler robots.txt allows, and an old directory listing whose outdated prices assistants keep repeating. Both are cheap to fix once found.

• Crawler access: robots.txt groups, CDN and firewall rules, bot challenges and the published IP ranges of each platform's crawler.
• Indexability: pages indexed in Google and Bing, canonicals, snippet controls and content that only exists in images or scripts.
• Entity clarity: one name, description, location and service list across your site, Google Business Profile, LinkedIn and directories, distinct from similarly named businesses.
• Answer pages: the questions buyers ask before contacting you, such as price drivers, process and fit, mapped to pages or marked as gaps.
• Outside sources: the directories, reviews, articles and forums assistants cite for your category, and where you are missing or misdescribed.
• Current answers: how each assistant describes you today, whether it is accurate and where each error comes from.

How do you measure visibility in AI answers?#

With several partial sources read together, because no single tool sees every assistant:

The panel is an observation, not a ranking. Google notes that no third-party tool has access to its internal ranking or AI systems, so we keep the raw answers rather than a vendor's score.

• Search Console's Generative AI performance report: AI Overviews and AI Mode impressions by page, country and device. Small sites may see no data; AI clicks also count under Web.
• Bing Webmaster Tools' AI Performance report, in public preview since February 2026: how often Copilot and Bing's AI summaries cite your pages, and which ones.
• Server logs: search crawlers that build each index, and user fetchers such as ChatGPT-User, Claude-User and Perplexity-User that read a page while answering someone.
• Analytics: visits from chatgpt.com, perplexity.ai and claude.ai, and whether they became inquiries.
• A fixed prompt panel: the same buyer questions in English and Spanish each month, recording mention, cited page, accuracy and competitors named.

Is your website ready for AI agents that act for buyers?#

Assistants are starting to do tasks for people: compare providers, read prices, check availability, start an inquiry. Google describes browser agents that read screenshots, page structure and the accessibility tree. Preparing for them is a separate job from search optimization, and none of the layers below is a ranking factor.

dardo.studio runs the full set, so we describe it from first-hand use. Every page has a markdown copy (add index.md or send Accept: text/markdown) with a canonical Link header back to the HTML page. llms.txt and llms-full.txt summarize the site in both languages. robots.txt names the search and user agents of OpenAI, Anthropic, Perplexity and Bing, with Content Signals allowing search and AI input. A public MCP server offers two read-only tools that list our services and prepare a project brief without sending it. The same tools answer over A2A, described by an agent card, and through WebMCP in browsers that support it, and an RFC 9727 API catalog lists the public endpoints.

Adoption is early. An analysis Cloudflare published in April 2026, covering the 200,000 most visited domains, found Content Signals on 4% of sites and markdown negotiation on 3.9%, with MCP Server Cards and API catalogs on fewer than 15 sites combined. A shop with live stock or bookable slots gives an agent something to do; a brochure site mostly needs clean HTML, labelled forms and accurate content. We tell you which layers your business needs.

Questions before you choose#