How well can A.I. find your site?
Your customers ask ChatGPT, Google and Perplexity who to trust. This free check scores any page 0–100 on whether those engines can fetch it, read it and cite it, and tells you exactly what to fix.
Free · No signup · Results in about 10 seconds · Shareable report
Presented by Nick Krem A.I. Strategist for real estate · nickkrem.com - OpenAI
- Anthropic
- Perplexity
- Meta AI
- Apple
The answer is the new front page.
A billion people a month ask Google's AI Mode instead of scanning ten links, and nearly as many use ChatGPT every week. When an engine picks one to three sources, it picks from pages it can fetch and parse. Outside Google's own surfaces, most cited pages never ranked on page one.
Five categories, weighted by what actually removes you from answers.
Fetchability carries the most because it is binary in practice: a page that returns a bot challenge to OAI-SearchBot is cited exactly as often as a blank page. Content heuristics carry the least because they are proxies, not measurements.
Can an AI crawler load the page at all: no WAF challenge, no 403, no JavaScript-only shell, and a robots.txt that does not lock out the agents that produce citations.
JSON-LD the engines read, an article body Mozilla Readability can extract, author and organization provenance, a fresh dateModified, and site depth.
The foundation every retrieval index was built on. Google's AI Overviews and AI Mode are ordinary search results underneath; this category is most of what decides them.
The outline a parser uses to separate the answer from the chrome.
The two effects from the Princeton GEO study that replicated, plus the answer-first structure that survives query fan-out.
Phone, address, hours and service area, from JSON-LD or visible text. Counted only when it applies, so a SaaS page is neither penalized nor rewarded.
Every point, and the reason it was or was not earned.
A result is a list, not a verdict. Each row names the check, the points, and the specific thing to change. Fix in weight order, re-run, and the permalink keeps every version.
Reachable and readable. Little for an engine to cite: nothing on the page names the agent, the market or the service in a form a machine can lift.
The same domain, the same 27 checks, before and after the findings were fixed.
The report is free and stays at its link. When the fixes are more than you want to take on, our team does them and re-runs this audit until the page is an A. Real-estate professionals can have Halo, the marketing agency you text, do it as part of their site.
Verified pair · www.shannonmiles.com · a 28th check (contact signals) was added after May and is excluded
Blocking the wrong one costs nothing. Blocking the right one removes you from the answer.
Every AI user agent does one of three jobs, and the operators document them differently. The check scores only the two classes whose blocking removes you from answers. The full 2026 reference →
Build the index answers are drawn from
Blocking one removes your pages from that engine's answers. OpenAI: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers."
- OAI-SearchBot · ChatGPT search
- Claude-SearchBot · Claude
- PerplexityBot · Perplexity
- Googlebot · AI Overviews, AI Mode
- Bingbot · Copilot
Load your page because a user just asked
Blocking one means the assistant tells the user it could not read your site. Most ignore robots.txt; only an edge rule stops them, and it stops the citation too.
- ChatGPT-User · ChatGPT
- Claude-User · Claude
- Perplexity-User · Perplexity
- Meta-ExternalFetcher · Meta AI
- Google-Extended · Gemini grounding
Collect text for future models
Blocking these costs no citations. Cloudflare blocks this class by default on ad-monetized pages for every zone onboarded from September 15, 2026.
- GPTBot · OpenAI
- ClaudeBot · Anthropic
- Meta-ExternalAgent · Meta
- Bytespider · ByteDance
- CCBot · Common Crawl
What people ask before they run the check.
What is Answer Engine Optimization?
AEO is structuring a website so that generative engines, ChatGPT, Claude, Perplexity, Google AI Overviews and AI Mode, can fetch it, parse it and cite it as a source. It overlaps with SEO on the foundations and adds signals engines read that Google historically did not: an allowlist for AI agents in robots.txt, JSON-LD the engines extract entities from, an article body a parser can lift cleanly, and answer-first content structure.
How does the score work?
27 weighted checks in five categories, plus one contact-signals check that applies only to businesses, listings and people. Points earned and possible are summed across the checks that ran and normalized to 0–100. A ≥ 90, B ≥ 80, C ≥ 70, D ≥ 60, F below.
Does it run my site through a headless browser?
No. The check reads the server-rendered HTML directly, which is what AI crawlers do. A page that needs JavaScript to show its content fails the server-rendered content check, on purpose.
Which AI crawlers does it check, and which ones matter?
Fifteen agents are read from robots.txt; seven are scored: OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User and Google-Extended. Training-only crawlers such as GPTBot, ClaudeBot and CCBot are reported but not scored.
One correction worth knowing: Google-Extended controls Gemini training and grounding only. Google AI Overviews and AI Mode are governed by ordinary Googlebot access and the standard snippet controls.
Do I need an llms.txt?
Not for citations. Ahrefs checked 137,000 domains in May 2026 and found 97% of published llms.txt files were never requested; SE Ranking found no correlation between having one and being cited across 300,000 domains, and Google says it does not read the file. Coding agents do request it and it costs nothing, so the check still awards points for a valid file.
What changes on September 15, 2026?
Cloudflare's defaults for newly onboarded domains block Training and Agent crawlers on pages that show ads, keep Search crawlers open, and stop treating "Verified" as "allowed". If your site is on Cloudflare, check the AI crawler policy for your zone; a Training block now also catches multi-purpose crawlers such as Googlebot on those pages.
Is the report really free? What happens with my details?
The full report is free, needs no signup, and stays at its link. If you ask us to text you a fix plan, we use your number for that report and the follow-up you asked for, nothing else. Having the fixes done is a separate, paid service, offered only after you have the free report.
Latest
Sixteen articles: primers, the rubric, and research summaries with the numbers checked against the primary sources.