Question 1

What is the AI Trust Audit?

Accepted Answer

The AI Trust Audit is a free analysis tool from Trusted Shops. It checks websites for over 70 factors in 7 categories (Crawling, Authority, Security, Compliance, Reputation, Content), providing a score of 0–100% with recommendations for action. Target audience: operators who want to be visible in ChatGPT, Perplexity, and Google Gemini.

Question 2

Who is the AI Trust Audit suitable for?

Accepted Answer

Target groups include website operators, SEO and marketing managers, as well as e-commerce shops, service providers, and content platforms. The audit shows how AI systems view a website and can be used for any industry.

Question 3

Is the audit really free?

Accepted Answer

No. The AI Trust Audit is free and can be used without registration. There is no subscription, no limits on the number of audits, and no paid premium tier.

Question 4

How do AI systems see my website?

Accepted Answer

AI crawlers experience your website very differently from a browser. A 2026 Writesonic study tested six major LLMs (ChatGPT, Claude, Gemini, Copilot, DeepSeek, Grok): three of them execute no JavaScript at all, the other three wait only 500 ms to 3 seconds. None of them scrolls — below-the-fold content and lazy-loaded images stay invisible. In direct conversation mode, all six LLMs ignore the entire section except the tag — including JSON-LD and meta description. What stays readable: the initial HTML, semantic headings, microdata in the body and <noscript> fallbacks. Source: writesonic.com/blog/ai-crawler-study-what-llms-see-on-your-website

Question 5

What is the difference between SEO and GEO?

Accepted Answer

SEO optimizes a website for classic search results (Google) via keywords, backlinks, and technical factors. GEO (Generative Engine Optimization) optimizes content so that AI systems like ChatGPT, Perplexity, or Google AI Overviews cite it as a source – focusing on structured data, entity authority, and machine-readable content. The AI Trust Audit covers both areas.

Question 6

Which AI crawlers exist and what do they do?

Accepted Answer

The most important AI crawlers are: GPTBot (OpenAI/ChatGPT), OAI-SearchBot, ChatGPT-User, PerplexityBot (Perplexity AI), ClaudeBot (Anthropic), Google-Extended (Google Gemini), Bytespider (ByteDance), Cohere-AI, and Meta-ExternalAgent. They crawl websites for model training or real-time answers. The robots.txt controls which bots have access.

Question 7

Is there a premium version of the audit?

Accepted Answer

No. The AI Trust Audit is completely free – with no registration required, no subscription, and no feature limitations. Trusted Shops offers separate, complementary consulting services.

Question 8

What are the advanced options in the input form for?

Accepted Answer

The advanced options (gear icon in the URL input field) allow you to manually set URLs for product pages, blog/guide pages, and FAQ pages. These URLs override the automatic detection from Sitemap and HTML. Use this if automatic detection selects the wrong subpages or certain pages are not in the Sitemap.

Question 9

How is the AI Trust Score calculated?

Accepted Answer

The score is based on over 70 automated checks in 7 categories (Crawling & Indexing, Authority & Structure, Security & Technology, Compliance & Legal, Reputation & Social Proof, Content Quality). Each check is weighted 0–5 points; checks with weight 0 are informative and do not contribute. The total score = points achieved / maximum possible points, presented as a percentage.

Question 10

What is a good score?

Accepted Answer

A score above 80% = very good. 70–80% = solid with optimization potential. 40–60% = basic implementation with significant gaps. Below 40% = urgent need for action. The score is an indicator; individual check results show specific optimization points.

Question 11

Why are some checks failed or skipped?

Accepted Answer

Checks fail due to bot protection (WAF), unreachable external APIs, or missing public data. Skipped checks are not relevant for the website type – e.g., Trusted Shops and Product-Schema for non-shop pages. Both statuses are marked in the results.

Question 12

What do the six audit categories mean?

Accepted Answer

(1) Crawling & Indexing: Is the website accessible to AI crawlers? (2) Authority & Structure: Is it recognized as a trustworthy source? (3) Security & Technology: HTTPS, Security Headers, Performance. (4) Compliance & Legal: Impressum, Privacy Policy, GTC, Cookie Consent Banner. (5) Reputation & Social Proof: external perception (Wikipedia, Reddit, reviews). (6) Content & Semantics: machine-readable, semantically structured content.

Question 13

Why do I get different results when analyzing the same website repeatedly?

Accepted Answer

Fluctuations of a few percentage points are normal. Causes: (1) Bot protection and WAFs react differently depending on IP and time of day. (2) CDN geo-routing delivers varying content. (3) Dynamic elements (cookie banners, A/B tests) change the page per request. (4) External APIs (Wikipedia, Reddit, Trusted Shops) are not always equally available. (5) The AI verifier is not completely deterministic. The overall trend remains reproducible.

Question 14

What are informational checks?

Accepted Answer

Informative checks have a weight of 0 and do not contribute to the score. They simply show whether a feature is present. Examples: HowTo-Schema, Video-Schema, Cookie-Banner, Wikipedia presence, comment function, FAQ-Schema, Event-Schema, Breadcrumb-Schema, Recipe-Schema. Reason: They are highly dependent on the website type or industry.

Question 15

Why do I see more points than the displayed maximum on some checks (e.g. "4/3 points")?

Accepted Answer

This is not a display error. When an audit runs, we permanently freeze the points you earned – your result stays comparable even when we refine the methodology later. The maximum per check, however, is shown based on the current weighting so you can see how the check would be scored today. If we lowered a check's weight after your audit (e.g. from 4 to 3 points), your historically awarded points can exceed today's maximum. In short: you achieved the full score in that check. A new audit would simply show the current maximum (e.g. 3/3) here.

Question 16

How does the technical analysis work?

Accepted Answer

The audit retrieves the HTML page, parses robots.txt and Sitemap, and automatically identifies representative subpages (product, blog, FAQ). It checks HTTPS headers, security settings, structured data (JSON-LD, Schema.org), meta tags, and HTML structure. External APIs provide reputation data (Trusted Shops, Wikipedia, Reddit). Schema checks use subpages if the homepage contains no suitable data.

Question 17

Why does the analysis sometimes take longer?

Accepted Answer

An analysis typically takes around 30 seconds, and exceptionally up to 2 minutes. Factors: website load time, response time of external APIs (Wikipedia, Reddit, Trusted Shops), HTML size, and current server load.

Question 18

Can the audit damage or overload my website?

Accepted Answer

No. The audit is read-only and sends a single HTTP request to the homepage plus one request each to detected subpages. No forms are filled out, no logins are performed, and no repeated crawls are initiated. The load is equivalent to a normal page view.

Question 19

Why does my website with JavaScript rendering (React, Vue, Angular) score lower?

Accepted Answer

Single Page Applications (React, Vue, Angular) load content client-side via JavaScript. The 2026 Writesonic study shows: ChatGPT, Claude and Gemini execute no JavaScript at all. Copilot, DeepSeek and Grok do render JS but wait only 500 ms to 3 seconds — lazy-loaded content, slow widgets and asynchronously fetched reviews stay invisible too. What is missing in the audit will most likely be missing for AI systems. Solution: Server-Side Rendering, pre-rendering or static generation so that main content, structured data and critical facts are already in the initial HTML.

Question 20

What is structured data and why does it matter for AI?

Accepted Answer

Structured data (JSON-LD per Schema.org) is machine-readable markup in your HTML. Examples: Organization, Product, FAQPage, Article, AggregateRating. Important: JSON-LD remains essential for Google discovery and the classic search index — but in direct LLM conversation mode, none of the six models tested in the 2026 Writesonic study reads it. Recommendation: keep JSON-LD in the <head> and additionally expose critical facts (prices, availability, ratings) as microdata in the visible body. That way your content is readable for both Google and ChatGPT, Perplexity and Claude.

Question 21

How does robots.txt affect AI visibility?

Accepted Answer

The robots.txt controls access for individual bots. For AI visibility, GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot, Google-Extended, Cohere-AI, and Meta-ExternalAgent must not be blocked. Blanket Disallow rules often unintentionally block AI bots. The audit checks all relevant crawlers individually and lists missing Allow rules.

Question 22

Which content stays truly invisible to AI crawlers?

Accepted Answer

The 2026 Writesonic study reveals four blind spots: (1) Lazy-loaded and below-the-fold content — no LLM crawler scrolls the page. (2) JavaScript-injected widgets without server-side rendering, such as embedded review modules or dynamic product data. (3) CSS-generated content via ::before and ::after — these pseudo-elements are ignored. (4) Anything behind login, cookies or geo-blocking. Tip: <noscript> fallbacks are read by 4 of 6 LLMs — critical content such as product descriptions or review text can be mirrored there when JS rendering is not an option.

Question 23

How can I improve my score?

Accepted Answer

The results page shows a specific recommendation for action for each failed check. Order: first, checks with high weighting (4–5 points), then quick wins like structured data, meta tags, and HTTPS optimization. Reputation and content topics follow, as they take longer.

Question 24

How fast do improvements take effect?

Accepted Answer

Technical changes (HTTPS, meta tags, JSON-LD) are visible during the next crawling – usually within hours to days. Reputation signals (Wikipedia, Trusted Shops, reviews) take weeks to months. Content improvements have a delayed effect. Recommendation: Re-audit after 1–2 weeks.

Question 25

Do I need technical knowledge for implementation?

Accepted Answer

Partially. Meta tags and structured data can be implemented via CMS or plugins. HTTPS configuration, security headers, and robots.txt require developer support. Each recommendation includes specific instructions with code examples.

Question 26

What are the most important quick wins for better AI visibility?

Accepted Answer

The five most important quick wins: (1) Enable HTTPS and set security headers (HSTS). (2) Integrate JSON-LD for Organization, Product, or FAQPage. (3) Open robots.txt for AI crawlers (GPTBot, PerplexityBot, ClaudeBot, Google-Extended). (4) Optimize Meta-Title (< 70 characters) and Meta-Description (< 160 characters) on every page. (5) Create an XML sitemap and reference it in robots.txt. Implementation time: usually a few hours.

Question 27

How often should I repeat the audit?

Accepted Answer

After every major website change, at least every 4–6 weeks. The AI landscape changes rapidly: new crawlers appear, algorithms are adjusted, external reputation signals are updated. Regular audits show progress and new optimization potentials.

Question 28

Can the recommended optimizations have negative effects on SEO?

Accepted Answer

No. All recommended measures are SEO-neutral or SEO-beneficial. (1) JSON-LD is embedded in invisible script tags – no layout effect. Google explicitly recommends JSON-LD. (2) JSON-LD enables rich results (stars, FAQ snippets, product info) and improves CTR. (3) Allow rules for AI crawlers have no impact on Google ranking, as Google uses its own Googlebot. (4) HTTPS, security headers, sitemap, meta tags are classic SEO Best Practices.

Question 29

How has the 2026 Writesonic study changed my optimization strategy?

Accepted Answer

Three clear consequences: (1) The <title> tag is today the most important head element — 5 of 6 LLMs read it, meta description and OG tags barely. Invest in precise, information-dense titles under 70 characters. (2) Above-the-fold HTML beats any lazy-load animation. Put USP, price, availability and primary CTA directly into server-rendered HTML. (3) Double up on structured data: keep JSON-LD for Google and additionally expose critical facts as microdata in the visible body. This way you reach both the classic search index and conversational search.

Question 30

What does 'Website not reachable' mean?

Accepted Answer

'Website unreachable' means the trial request did not receive an HTML response. Common causes: misspelled URL, website offline, bot protection (Cloudflare, WAF) blocking the request, geographic access restriction, or temporary server issue. Check the URL, temporarily disable bot protection, or try again later.

Question 31

Why are many checks shown as 'Technical error'?

Accepted Answer

Multiple checks with 'Technical Error' indicate aggressive bot protection, geo-blocking, or unreachable external APIs (Wikipedia, Reddit). Successfully completed checks remain meaningful; however, the overall picture is incomplete. Workaround: Temporarily loosen bot protection and repeat the audit.

Question 32

My website has a low score – is that bad?

Accepted Answer

No. A low score indicates optimization potential – the majority of websites are not yet optimized for AI systems. Use the results as a roadmap: implement high-weighted checks first, then improve iteratively.

Question 33

Is my data stored during the audit?

Accepted Answer

Yes. The audit exclusively saves publicly accessible website data and analysis results, so the results page remains permanently accessible. With optional registration for the complete overview, contact data is processed in compliance with GDPR and used only for the specified purpose. Details: Privacy Policy.

Frequently Asked Questions

General Questions