---
title: "How to run an AI search audit: technical checklist | GEO Repair"
description: "A step-by-step audit process for measuring AI search readiness, cited sources, competitor presence, and the fixes your own site can control."
source: https://geo.repair/blog/how-to-run-ai-search-audit
last_updated: "June 18, 2026"
author: "GEO Repair Editorial"
reviewed_by: "GEO Repair Technical Review"
---

# How to run an AI search audit

> A step-by-step audit process for measuring AI search readiness, cited sources, competitor presence, and the fixes your own site can control.

**June 18, 2026** · AI Search, Audit, Measurement · By GEO Repair

An AI search audit measures how your brand and website appear in AI-driven discovery, then separates the problems you can fix on your own site from the problems that depend on third-party sources. The useful output is not a vague score. It is a list of pages, prompts, evidence, and fixes.

Start with a small scope. Pick the pages and questions that matter most, then audit them the same way every month.

## 1. Choose the pages and prompts

Audit the pages that influence buying decisions:

- Homepage
- Product or service page
- Pricing page
- Security or trust page
- Top educational guides
- Comparison or alternative pages, if you have them

Then choose 10 to 20 prompts a buyer might ask. Mix branded, category, and problem-based prompts:

- "best tools to improve AI search visibility"
- "how do I make my website show up in ChatGPT"
- "AI search optimization for SaaS websites"
- "is [brand] good for [use case]"
- "[brand] alternatives"

Keep the prompt list stable so you can compare results over time.

## 2. Check the technical page surface

For each target page, run the [AI search readiness checklist](/blog/ai-search-readiness-checklist). At minimum, record whether:

- Raw HTML includes the main content
- Metadata matches the page
- The canonical URL is correct
- Valid JSON-LD is present
- `robots.txt` and `sitemap.xml` are intentional
- The page has one `h1` and a clean heading order
- The page answers its main question directly

Save the evidence. A screenshot is less useful than the actual missing HTML, invalid schema error, or blocked crawler rule.

## 3. Test AI answers and cited sources

Run the prompt set in the AI search systems that matter to your buyers. For most teams, start with ChatGPT search, Perplexity, Google AI search, and Claude search if it is available in your region.

Track these fields:

- Prompt
- Platform
- Your brand mentioned: yes or no
- Your URL cited: yes or no
- Competitors mentioned
- Sources cited
- Accuracy of the answer
- Notes on missing or misleading claims

Do not overread one run. AI answers vary. The point is to find repeated patterns, not to treat one answer as a permanent ranking.

## 4. Separate site-side blockers from placement gaps

This is the step most audits skip.

A site-side blocker is something you can fix in your own website:

- The page is blocked or missing from the sitemap.
- Important copy is not present in raw HTML.
- The page title is generic.
- JSON-LD is missing or wrong.
- The answer is buried below filler copy.

A placement gap means the AI system is relying on external pages where you are absent:

- Third-party roundups
- Review sites
- Reddit threads
- YouTube videos
- Analyst or industry guides
- Competitor comparison pages

GEO Repair fixes the first group. The second group needs distribution, partnerships, reviews, and off-site work.

## 5. Create the fix queue

Turn the audit into a repair queue. Each item should have:

- Page
- Finding
- Evidence
- Why it matters
- Fix owner
- Fix type
- Re-check method

For example:

| Page            | Finding                            | Evidence                                   | Fix                                        |
| --------------- | ---------------------------------- | ------------------------------------------ | ------------------------------------------ |
| `/pricing`      | Pricing copy missing from raw HTML | `curl` output lacks plan names             | Server-render pricing content              |
| `/blog/example` | Article schema missing             | Rich Results Test finds no Article JSON-LD | Generate Article schema from post metadata |
| `/security`     | Generic title                      | Title says "GEO Repair" only               | Add page-specific title and description    |

## 6. Re-check after the fix

After changes ship, run the same checks again. The goal is not to prove that AI systems will cite you immediately. The goal is to prove that the blocker was removed.

Good after-fix evidence looks like this:

- Raw HTML now includes the missing content.
- JSON-LD validates and matches the page.
- The sitemap includes the URL.
- The answer appears near the top of the section.
- The audit row moves from failed to passed.

That is the part you control. Mentions, citations, and referral traffic should be tracked over time, but they are not instant proof of technical readiness.

## Use the audit to decide what to fix

If the audit finds crawler or rendering problems, fix those first. If the page is technically clean but absent from AI answers, look at answer quality, sources, and off-site placement. If competitors are cited from third-party lists, your website may not be the only surface that needs work.

For the source-level repair path, read [how to fix AI search visibility issues](/blog/fix-ai-search-visibility-issues) or run the [free AI search readiness scan](/#checkup).

---

_Markdown copy of [How to run an AI search audit: technical checklist | GEO Repair](https://geo.repair/blog/how-to-run-ai-search-audit), a faithful text version of the page for machines and readers. © GEO Repair._