How we check websites and write our guides

This page explains how we audit websites, prepare our guides and back up what we claim. It also says where our work stops: what it can establish and what it cannot.

What an audit checks

We request pages without executing JavaScript and look for the main text in the server response. That tells us the text is reachable without scripts. On its own it does not confirm that a real search crawler received the same response.

The site should use one preferred hostname, with or without www, and the other version should redirect to it permanently, with no chains or loops. To catch inconsistencies, we compare the URLs in canonical tags, hreflang annotations, the sitemap and llms.txt.

In robots.txt, we read the rules for GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended and Bingbot one by one. Schema.org markup is compared with the visible text, and we measure server response time and page size.

Then comes the content. Do the opening paragraphs answer the buyer’s question? Does the page give facts someone can verify, such as prices, timelines, service details, names and dates? The company description also has to be consistent across language versions.

How we measure AI answers

One answer proves little, because results change from one conversation to the next. We ask each question at least three times, starting a new conversation each time, and test each language and assistant separately. We record the date, whether the site is mentioned, where it appears in any list, which pages are linked and which other businesses are named. Reports include screenshots only when the answer has been checked again.

How we verify sources

Every figure taken from external material carries a source and a date. We keep primary sources, secondary reports and our own assumptions apart. Primary sources include Google, OpenAI or Anthropic documentation and published studies that describe their samples. Anything from a secondary report, or anything we are assuming, is labeled as such or left out. We compare assistant speed and answer quality only when we have measured them ourselves, and we do not apply findings from one market to another.

What we do not promise

We cannot guarantee whether, where or when a site will appear in AI answers. Technical fixes help systems access information, but which sources to cite is each service’s own decision. If part of a site already works correctly, we say so in the report.

How guides are written and updated

Before publication, we check each guide against its sources and the standards described here. The original publication date stays the same. We add an “Updated” note only when the content changes substantially, for example to reflect new data, changed platform rules or a factual correction. Fixing typos, links or layout does not change the date.

Found an error?

Write to hello@geo-rank.ai with the page URL and a description of the error. If we confirm it, we correct the guide and note the date of the correction.

Published: · geo-rank.ai