Generative engine optimization starts with a simple question: can the systems behind AI assistants actually read your website? Many SaaS sites are built as JavaScript applications that send an almost empty HTML page and draw the content in the browser. Some block AI crawlers in robots.txt without realizing it. Others are readable but never clearly explain what the product does, who it is for or what it costs. The GEO audit checks all of this from the public HTML of your site.
Crawler access
The audit reads your robots.txt and evaluates it for the crawlers that matter for AI visibility and search: GPTBot and OAI-SearchBot for OpenAI, ChatGPT-User for browsing, ClaudeBot and Claude-SearchBot for Anthropic, PerplexityBot, Google-Extended, Googlebot and Bingbot. For each one you see whether your homepage is allowed. It also checks your sitemaps and counts the URLs they list.
Allowed access is a requirement, not a guarantee. The audit states explicitly that an accessible page is not necessarily indexed, and it lists the checks it cannot perform from public HTML — such as Google index status, search impressions or field performance data — instead of guessing them. It also notes that llms.txt is not required for AI visibility.
Page-level checks
Using the pages Narrashift already analyzed for your project, the audit fetches each page and checks the HTTP status and redirects, the title and meta description and their length, duplicate titles and descriptions across pages, the number of H1 headings, the canonical URL, a noindex directive, JSON-LD structured data types, internal links and the amount of readable text.
One check deserves special attention for SaaS websites: if a page contains very little text in its HTML while loading several scripts, the audit flags that the content probably needs JavaScript to render. Many AI crawlers do not execute JavaScript, so such a page can look empty to them even though it looks complete in a browser. Server-side rendering or prerendering usually solves this.
Content coverage
Beyond technical signals, the audit checks whether your analyzed pages explain the topics buyers expect from a SaaS website: what the product does, who it is for, which problems it solves, its features and integrations, pricing, limitations, differentiators, use cases, comparisons with alternatives and frequently asked questions. Each topic shows whether it was found and on which page.
The audit also links to content-gap opportunities: monitored topics where AI answers leave you out and none of your pages covers the subject. Detection is keyword-based and the audit says so, so you can review the pages before acting.
Safe by design
Audits use the same protected crawler as the rest of Narrashift: only public HTTP and HTTPS addresses, no private networks, size and time limits, and a clear user agent. Up to twenty pages and a sample of internal links are checked per audit, and audits run in the background so your dashboard stays fast. The number of audits per month depends on your plan.
What to fix first
Start with anything marked high: blocked AI crawlers, pages that return errors, noindex directives and content that only appears with JavaScript. These decide whether AI systems can see your website at all. Then work on medium issues such as missing descriptions, missing H1 headings and broken internal links, and finally on the content topics your site does not yet explain. Re-run the audit after each round of changes to confirm the fixes are live.