How to Audit Your AI Search Visibility: A Practical GEO Testing Guide

You published your llms.txt. You front-loaded definitive answers. You added structured data. Now the question is: is any of it working?

Traditional SEO has Google Search Console — type in a URL, see impressions, clicks, position. GEO has nothing that simple. No dashboard shows “your site was cited 12 times by ChatGPT this week.” You have to test manually.

This guide walks through a practical, repeatable process for auditing your AI search visibility. No paid tools. No guesswork.

Why Manual Testing Beats Tools

Several “AI visibility tracking” tools launched in 2025-2026. Most run a fixed set of queries against one AI engine and sell you a score. The problems:

• They test a tiny sample of queries, not yours
• They can’t access personalized AI results
• They lag behind model updates (a tool calibrated for GPT-4 may be meaningless for GPT-5)
• They cost $50-$200/month for data you can gather yourself in 20 minutes

Manual testing is slower but more accurate. You test the exact queries your audience uses, on the exact platforms they use, with current models.

The 5-Platform Test

Test on these five AI search platforms. Each has a different retrieval system, citation format, and audience:

1. ChatGPT (with web search) — largest user base, uses Bing for retrieval
2. Perplexity — dedicated AI search, most transparent about sources
3. Google AI Overviews — appears at the top of Google search for many queries
4. Claude (with web search) — growing user base, uses its own retrieval
5. Microsoft Copilot — enterprise audience, Bing-based

You don’t need to test all five every time. Start with ChatGPT and Perplexity — they’re the most likely to cite independent sites. Add Google AI Overviews for queries where you already rank in traditional search.

Step 1: Build Your Test Query List

Don’t test random queries. Test the ones that matter to your site.

Pull from Google Search Console. Export your top 50 queries by impressions from the last 90 days. These are queries Google already associates with your site — the same queries AI engines are likely processing. As we covered in our GSC guide, sort by high impressions and filter for informational intent (questions, “what is,” “how to”).

Add question-format queries. AI search users type full questions, not keyword fragments. Transform your top keywords:

• “keyword research” → “how to do keyword research without paid tools”
• “SEO audit” → “what does a technical SEO audit check”
• “GEO optimization” → “how do I get my website cited by AI search”

Add comparison queries. AI engines handle comparisons well, and being cited in a comparison positions your brand:

• “what’s the best free SEO tool for indie developers”
• “Ahrefs vs free alternatives for keyword research”
• “ZensInk vs paid SEO tools”

Aim for 20-30 test queries covering your top topics.

Step 2: Run the Tests

For each query, type it into the AI search engine exactly as a user would. Record:

• Cited? (Yes/No) — Did your site appear as a source?
• Position — First citation? Second? Fifth?
• Quoted text — What did the AI extract from your content?
• Competitors cited — Which other sites appeared?

ChatGPT test:

Type your query in ChatGPT with web search enabled.
Look for source citations (numbered references or inline links).
Record results.

Perplexity test:

Type your query in Perplexity.
Check the "Sources" section below the answer.
Perplexity shows all sources — easy to scan.

Google AI Overviews test:

Search your query on Google.
Look for the AI Overview block at the top.
Check if your site is cited (click "more" to expand sources).

Pro tip: Test in incognito/private mode to reduce personalization bias. Run each query 2-3 times on different days — AI responses vary between runs.

Step 3: Score Your Visibility

Create a simple scoring system:

Citation rate: What percentage of your test queries cited your site? Example: 6 out of 25 queries = 24% citation rate.

Citation position: When cited, where do you rank among sources? First source is best — users click it most.

Coverage by platform: Are you cited on ChatGPT but not Perplexity? Or Google AI Overviews but not ChatGPT? Platform gaps reveal where to focus.

Benchmark: Run the same queries with your competitors’ brand names. If a competitor is cited for queries where you’re not, analyze why — is their content more structured? Do they answer more directly?

Step 4: Analyze Why You’re Not Cited

If you’re not being cited, the problem falls into one of three buckets:

Bucket 1: Not Retrieved (Indexing Problem)

The AI engine hasn’t discovered your content. This is a traditional SEO problem — if Google can’t find your page, neither can AI search engines.

Check:

• Is the page indexed in Google Search Console?
• Is the URL in your sitemap?
• Is the page blocked by robots.txt or noindex?
• Does the page have internal links pointing to it? Internal linking is how all crawlers — including AI — discover content.

Fix: Ensure the page is crawlable, indexed, and linked from at least 3 other pages on your site.

Bucket 2: Retrieved But Not Selected (Content Problem)

The AI found your page but chose a competitor’s content instead. This means your content exists but isn’t the best extractable answer.

Check:

• Does your page front-load the answer, or bury it in paragraph 8?
• Do you have structured data (tables, lists, schema)?
• Is your content quotable — clear, definitive, self-contained statements?
• Is the content original, or does it rephrase what competitors already say?

Fix: Rewrite the key section using BLUF structure. Add data tables. Make claims verifiable with sources. These same optimizations also improve traditional SEO rankings.

Bucket 3: Selected But Not Cited (Model Problem)

The AI may have used your content without citing it. This is harder to detect and harder to fix. It’s a model behavior issue, not a content issue.

Check:

• Does the AI’s answer contain information unique to your site (original data, proprietary research)?
• Does the AI paraphrase your exact phrasing without linking?

Fix: There’s no reliable fix for uncited use. The best defense is publishing original data and analysis that’s distinctive enough that any accurate answer must come from you. The more unique your content, the more likely AI engines cite rather than paraphrase.

Step 5: Track AI Referral Traffic

While manual testing shows citation rate, analytics show actual traffic impact.

In Google Analytics 4:

• Go to Reports → Acquisition → Traffic Acquisition
• Filter for source/medium containing: chatgpt, perplexity, copilot, you.com, claude.ai
• Set up a custom segment for AI referral traffic

In Umami or Plausible:

• Filter referrer URLs for AI domains
• Track month-over-month growth

AI referral traffic is still small for most sites — single digits as a percentage of total traffic. But it’s growing fast. Sites with strong GEO optimization report AI referrals doubling quarter over quarter.

Important: Don’t expect AI referral traffic to match citation rate. Many users read the AI’s answer (which cites you) but never click through. The value of being cited is brand visibility and authority, not just clicks.

Step 6: Set Up a Repeatable Audit Cadence

GEO testing isn’t one-and-done. AI models update frequently, citation behavior shifts, and competitors publish new content.

Monthly audit (30 minutes):

• Re-run your top 10 queries on ChatGPT and Perplexity
• Check AI referral traffic trends
• Note any new competitors appearing as citations

Quarterly audit (2 hours):

• Run the full 20-30 query list across all 5 platforms
• Re-score citation rate, position, and coverage
• Update your test query list based on new GSC data
• Compare against previous quarter’s scores

After major model updates:

• Test within 48 hours of any announced GPT/Claude/Gemini update
• Citation patterns can shift dramatically when models change

What Good Looks Like

Based on testing across multiple sites, here are realistic benchmarks for a site with solid GEO optimization:

Metric	Beginner	Intermediate	Advanced
Citation rate (your queries)	0-10%	15-30%	35%+
Average citation position	Not cited	3rd-5th source	1st-2nd source
Platforms with citations	0-1	2-3	4-5
AI referral traffic/month	0-10 visits	50-200 visits	500+ visits
Citation consistency	Random	Improving	Stable across model updates

These numbers are early-stage benchmarks. As AI search grows, expectations will rise.

Common Pitfalls

Testing too few queries. Testing 5 queries gives you no signal — AI responses are noisy. Test at least 20 queries, multiple times, before drawing conclusions.

Ignoring Google AI Overviews. Many developers focus on ChatGPT and Perplexity but overlook Google AI Overviews. Google’s AI answers appear above traditional results and reach the largest audience. If you rank in Google’s top 10 for a query, you’re a candidate for AI Overview citation.

Confusing mentions with citations. If ChatGPT mentions your brand name in its answer without linking to your site, that’s brand awareness — not GEO success. Track both, but prioritize earning actual citations with links.

Giving up too early. GEO is newer than SEO and noisier. A site that gets zero citations in month 1 might get 15% by month 3 as content gets indexed and models learn. Consistency matters more than early results.

FAQ

How often should I test my AI search visibility?

Monthly for a quick check (top 10 queries, ChatGPT + Perplexity), quarterly for a full audit (all queries, all platforms). Also test after major AI model updates.

What if my site is never cited by any AI engine?

Start by checking if your content is indexed and crawlable (traditional SEO). Then improve content structure — BLUF format, structured data, quotable statements. Run a free SEO audit to identify technical issues that block AI crawlers.

Do I need to pay for AI visibility tracking tools?

No. Manual testing is more accurate than paid tools because you test your actual queries on current models. Paid tools are useful if you manage many sites or need automated reporting, but for indie builders, 30 minutes of manual testing monthly is sufficient.

Does appearing in Google AI Overviews hurt my regular search traffic?

There’s no clear evidence that AI Overview citations cannibalize regular clicks. Some studies suggest sites cited in AI Overviews get more total clicks (traditional + AI), not fewer. Focus on being cited — the traffic math works out.

---

Start with a keyword research workflow to identify the queries your audience asks AI engines. Then apply GEO optimization strategies to make your content citable. Finally, audit your visibility monthly to track progress and catch gaps early.