# AI Search Audit Methodology: Rules and Scoring Understand the free audit’s crawl access, content discovery and answer structure checks, scoring weights and limitations. ## What this audit measures The audit makes read-only requests to public pages, robots.txt, sitemaps and selected URLs. It applies published technical rules without calling an AI model or changing your website. This report records technical readiness at the time of inspection and proposes work to do. The score is not a probability of AI citation. Requests using a crawler User-Agent do not prove that an official crawler visited. Indexing, rankings, citations and referral traffic require separate evidence. ## How the score works Crawl access has a 45-point pool, content discovery 25 and answer structure 30. Applicable checks are weighted within each pool. Partial matches receive partial credit; non-applicable checks are excluded. Optional signals can add up to five points, with the final score bounded at 100. Access and indexing problems may cap the total to make important issues visible. The published caps are 29 for a non-200 public page; 39 for noindex, a robots.txt block on OAI-SearchBot or a restrictive Content-Signal preference; 49 for a blocked OAI-SearchBot test request; and 59 for a robots.txt block on Bingbot. These are MadeInRed’s priorities, not a scoring system from Google or OpenAI. ## Business context Product and purchase signals activate ecommerce checks. Languages, currencies, market references and delivery information can activate cross-market checks. These examine visible facts and discoverability; they do not infer sales, market demand or platform endorsements. ## Crawl access · 45 points Check Weight What to review OAI-SearchBot access in robots.txt robots-oai-searchbot 10 Review the robots.txt group that applies to OAI-SearchBot and allow the public pages you want search to discover. This is separate from training access. ChatGPT-User policy in robots.txt robots-chatgpt-user 5 Review the declared ChatGPT-User policy and test public page access. User-initiated access has different controls from search crawling. Bingbot access in robots.txt robots-bingbot 5 Review Bingbot rules for the affected URLs. Allow relevant public content and verify real requests in server logs. Googlebot access in robots.txt robots-googlebot 2 Review Googlebot access for the pages intended for Google Search. Other AI search crawler policies robots-ai-answer-engines 3 Review the applicable crawler groups for the answer engines relevant to your audience. Content-Signal search and AI-input preferences content-signal 3 Review search and ai-input declarations in robots.txt or the CDN configuration. Set these preferences to match your publishing policy. HTTP access with an OAI-SearchBot User-Agent waf-oai-searchbot 9 Review the recorded HTTP response and CDN rules. Retest the URL with the stated User-Agent and verify actual official crawler visits separately. HTTP access with a ChatGPT-User User-Agent waf-chatgpt-user 5 Check the HTTP response, challenges and access rules. A simulated User-Agent is not proof of an official visit. HTTP access with a Bingbot User-Agent waf-bingbot 4 Check the HTTP response and any firewall challenge. Use server logs and Bing Webmaster Tools to confirm genuine crawler access. Page indexing directives indexable 5 Review robots meta tags and X-Robots-Tag response headers. Remove unintended noindex directives from pages intended for search. Snippet permissions snippet-allowed 3 Review nosnippet and max-snippet directives. Allow excerpts where consistent with your publishing policy. HTTP and HTTPS health http-health 3 Check the final status, redirect chain and response time. Serve the intended public content over HTTPS with a successful status. Real 404 responses for missing pages soft-404 2 Return a real 404 or 410 for missing content and verify an invented URL does not return an ordinary successful page. GPTBot training policy robots-gptbot Info Choose your training policy independently of search access. This informational check does not measure citations. CDN and firewall identification waf-detected Info Use the detected provider as a starting point for reviewing crawler access. Provider detection alone does not identify a blocking rule. ## Content discovery · 25 points Check Weight What to review Sitemap presence and robots.txt declaration sitemap-present 6 Publish a valid sitemap and reference its URL in robots.txt. Include the canonical pages you want discovered. Sitemap modification dates sitemap-freshness 4 Use accurate lastmod values for material updates. Avoid changing dates without a corresponding content update. Sampled sitemap URL health sitemap-urls-alive 5 Review the sampled URLs. Replace dead links and redirect entries with their final canonical URLs. Public Bing Webmaster Tools verification signals bing-verification 5 Confirm property ownership in Bing Webmaster Tools. DNS verification may not be visible in public HTML. Public Search Console verification signals google-verification 1 Confirm property ownership in Search Console. Absence of a public verification tag does not rule out DNS verification. Canonical URL consistency canonical 3 Declare a stable canonical URL and align internal links, sitemap entries and redirects with it. Language alternatives hreflang 2 For multilingual content, declare valid reciprocal hreflang links and correct language URLs. Single-language sites may not need this. Cross-market buying information export-market-signals 3 State languages, target markets, currencies, delivery areas and lead times. Provide accessible company, contact and policy pages. Discoverable RSS or Atom feed feed 1 If you publish regularly, provide a valid feed and expose its discovery link in the page head. HTTP to HTTPS redirect https-redirect 2 Redirect HTTP requests to the intended HTTPS URL without unnecessary hops. Page title and meta description title-description 2 Write a specific title and a useful standalone description that match the page’s primary purpose. IndexNow submission readiness indexnow Info Review your IndexNow key and submission records in your own configuration. Public inspection cannot confirm a private submission history. ## Answer structure · 30 points Check Weight What to review Readable content in the initial HTML ssr-content 6 Include the main answer and essential facts in the HTML response so they are available without running page scripts. Heading hierarchy heading-structure 3 Use a clear primary heading and descriptive section headings. Review skipped levels and unrelated sections. Headings aligned with reader questions question-headings 4 Use headings that help readers locate their questions and the relevant answers. Question phrasing is optional. Direct answer passages direct-answers 4 Place a useful answer near the relevant heading. Passage length is a local heuristic, not an official citation rule. Content structure on sampled pages sampled-pages-shape 5 Review the sampled product, article or category pages individually. Check readable answers, headings and factual completeness. Product and Offer structured data product-schema 3 Keep Product and Offer facts consistent with visible product information, including price, currency, availability and applicable identifiers. Product and purchase facts commerce-content 2 State the product, price and currency, availability, purchase action, and delivery or returns information. Keep these facts visible and consistent with JSON-LD. Accessible company, contact and policy pages trust-pages 2 Provide public company, contact and relevant policy pages and link to them clearly. Verify their final HTTP responses. Machine-readable publication and update dates dates 3 State accurate dates where relevant and keep visible dates consistent with structured data. Author or publisher attribution author 2 Identify the responsible author or publisher in visible content and appropriate structured data. JSON-LD structured data structured-data 3 Use valid structured data that describes the actual page. Do not add facts that are absent from visible content. Organization identity and sameAs links org-entity 2 Identify your organization consistently and link genuine official profiles with appropriate sameAs values. Useful lists and tables extractable-structures 2 Organise steps and comparisons into semantic lists or tables where they improve the answer. Links to supporting sources outbound-citations 2 Link factual claims to relevant, verifiable sources. Explain the scope of data and observations. HTML language declaration lang-declared 1 Declare the page’s actual language in the html lang attribute. Open Graph preview metadata open-graph 1 Provide a suitable og:title, og:description and og:image consistent with the page. Main content landmarks main-landmark 1 Use main or article elements to distinguish substantive content from navigation. Image alternative text image-alt 1 Add descriptive alt text to informative images. Decorative images may have an empty alt attribute. ## Additional signals · +5 points Check Weight What to review llms.txt site guide llms-txt 2 Consider a concise machine-readable guide to important public content. This is optional and is not a citation requirement. llms-full.txt content guide llms-full-txt 1 Consider a fuller text guide if useful to your audience. This optional convention is not a ranking guarantee. Markdown content negotiation markdown-negotiation 1.5 If supported, return a faithful Markdown representation for Accept: text/markdown while keeping the same access rules. Machine-readable API catalogue api-catalog 0.5 Document public API endpoints in a catalogue if you provide an API. Review and AggregateRating data review-schema 1 Use review or rating markup only for genuine, visible reviews and ratings that meet the applicable requirements. FAQPage or HowTo data faq-schema 1 Use structured data only when it accurately reflects visible questions, answers or instructions. This does not guarantee a rich result or AI citation. Official social and community sources source-ecosystem 1 Link active, relevant official video and community profiles. A profile link does not measure a citation or prove an endorsement. Conditional HTTP response headers conditional-headers 0.5 Use ETag or Last-Modified appropriately to support efficient conditional requests. ## Official references [OpenAI crawler documentation](https://platform.openai.com/docs/bots): search, user-initiated access and training are distinct. [Google crawler documentation](https://developers.google.com/search/docs/crawling-indexing/overview-google-crawlers). [Bing Webmaster Tools documentation](https://www.bing.com/webmasters/help/). [Schema.org](https://schema.org/): vocabulary used for structured data. Page content may quote its source language. Observations are limited to the time and URLs inspected. No score promises indexing, citation or ranking. Source: https://madeinred.com/en/methodology