How to Audit Whether Your Brand Appears in AI Answers

If buyers research your category in ChatGPT, Perplexity, or Google AI experiences and your brand never appears, you are losing consideration before the click. The practical way to diagnose that gap is an AI Answer Visibility Audit: confirm you are eligible to be retrieved, probe the questions that matter, score mentions versus citations, then decide whether to fix access, strengthen source pages, or invest in generative engine optimization (GEO).

This article shows how to run that audit in one working week, interpret the results without inventing vanity metrics, and choose the next move with evidence.

Key takeaways

• For Google AI Overviews and AI Mode, pages must be indexed and snippet-eligible; Google states there are no extra technical requirements beyond Search fundamentals.

• For ChatGPT search, OpenAI emphasizes allowing OAI-SearchBot and notes that placement is not guaranteed.

• Treat probe tests as directional evidence. Pair them with Search Console generative AI impressions when available, plus referral signals such as utm_source=chatgpt.com when ChatGPT sends traffic.

• Fix eligibility and extractable, non-commodity source pages before buying “AI ranking” tactics that ignore crawl access and content quality.

What “appearing in AI answers” actually means

Businesses often mix three different outcomes:

1. Brand mention — the answer names your company without linking to you.

2. Supporting citation / link — the answer or source panel links to a page on your site.

3. Preferred recommendation — the model positions you as a shortlist option for a buying question.

Those outcomes are not equal. A mention without a clickable path can still shape consideration, but a citation is usually more useful for measurement and traffic. A preferred recommendation is rarer and typically depends on clear entity signals, corroboration across the web, and pages that answer commercial questions directly.

Google’s documentation frames AI Overviews and AI Mode as Search features that ground responses in the Search index and surface supporting links. OpenAI’s publisher guidance likewise focuses on crawl access so content can be discovered, summarized, and cited — not on a guaranteed “rank #1 in ChatGPT” mechanic.

The AI Answer Visibility Audit

Score each dimension from 0–2 (0 = blocked or missing, 1 = partial, 2 = solid). Total possible: 12.

Interpretation: 0–4 means eligibility or fundamentals are blocking progress. 5–8 means you are partially visible and should prioritize source-page quality on commercial prompts. 9–12 means the issue is usually competitive depth and corroboration, not basic access — a stronger case for sustained GEO and authority work.

Step 1: Confirm eligibility before you blame “AI SEO”

Google’s AI features documentation is direct: to be eligible as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to appear in Google Search with a snippet. There are no additional technical requirements beyond Search technical requirements and policies.

Check these first:

• robots.txt and CDN rules allow Googlebot on the pages that should compete.

• Important content exists as text, not only inside images or inaccessible widgets.

• Preview controls such as nosnippet, aggressive max-snippet, or noindex are not silently excluding the pages you care about.

• Structured data, if used, matches visible page content.

For ChatGPT search, OpenAI states that ranking factors are designed to surface reliable information and that placement is not guaranteed. Inclusion depends in part on allowing OAI-SearchBot to crawl and on hosts/CDNs permitting OpenAI’s published IP ranges. Publisher guidance also notes referral tracking via utm_source=chatgpt.com when ChatGPT sends traffic.

If eligibility fails, stop debating GEO tactics and fix access. Visibility work cannot compensate for pages that systems cannot retrieve as snippet-eligible sources.

Step 2: Build a probe set that matches real research behavior

Create a fixed list of 20–40 prompts. Keep the set stable so month-to-month changes are comparable. Use four buckets:

1. Brand probes — “[Brand] reviews,” “[Brand] pricing,” “is [Brand] legit.”

2. Category probes — “best [service] for [audience] in [city/region],” “how to choose a [category] provider.”

3. Comparison probes — “[Brand] vs [Competitor],” “[option A] vs [option B] for [use case].”

4. Problem probes — the jobs buyers ask before they name a vendor (“why is my CPA rising after tracking changes,” “theme vs headless Shopify for a catalog refresh”).

Run the same prompts in Google Search surfaces where AI features appear, plus at least one assistant with web search citations (for example ChatGPT search). Log date, surface, whether you were mentioned, whether you were linked, which URL was cited, and which competitors appeared.

Do not treat one lucky prompt as proof of market visibility. Look for patterns across the set.

Step 3: Score citation quality, not just presence

For each prompt, assign the highest outcome you earned:

• 0 — absent

• 1 — mentioned without a useful link

• 2 — cited/linked

• 3 — recommended as a shortlist option with a clear reason

Then calculate two simple rates:

• Visibility rate = prompts with score ≥1 / total prompts

• Citation rate = prompts with score ≥2 / total prompts

A brand with a high visibility rate and low citation rate often has entity awareness but weak source pages. A brand with strong citations on branded probes and near-zero category visibility usually needs better commercial content and corroboration, not another homepage rewrite.

Step 4: Inspect whether your pages are extractable sources

Google’s generative AI optimization guidance emphasizes unique, non-commodity, people-first content and clear technical structure. It also myth-busts several popular shortcuts: special “AI-only” files are not required for Google Search AI features, chunking content purely for machines is unnecessary, and inauthentic mention campaigns are not a substitute for quality.

For each priority URL that should win citations, ask:

• Does the page answer the likely question in the first screen of text?

• Would that passage still make sense if lifted out of the page?

• Does it add first-hand process, data, constraints, or decision criteria competitors lack?

• Is the main content easy to distinguish from chrome, pop-ups, and boilerplate?

If competitors are cited and you are not, open their cited URLs. Compare extractability and uniqueness before assuming you need more pages.

Step 5: Connect probes to measurement systems

Manual probes are useful, but they are still samples. Where available, use Google Search Console’s generative AI performance report to see impressions from AI Overviews and AI Mode by page, country, and device. Google notes the report is rolling out and may be unavailable when a property has too few generative AI impressions or lacks access during rollout.

Also review analytics for referral patterns. OpenAI documents that ChatGPT can append utm_source=chatgpt.com on referrals, which makes assistant-driven sessions easier to isolate than generic “direct” traffic.

Use measurement to answer two questions: Which pages already earn AI-feature impressions? Which high-value prompts still produce zero citations after eligibility is clean?

Worked example: regional home-services brand (hypothetical)

Imagine a multi-location HVAC company with solid classic SEO for “furnace repair [city]” but almost no presence when buyers ask assistants for “best HVAC company for heat-pump retrofit near me.”

A one-week audit finds:

• Googlebot access is fine, but OAI-SearchBot is blocked by a CDN bot rule → Crawl eligibility = 1.

• Service pages are indexed and snippet-eligible → Snippet readiness = 2.

• Brand probes often mention the company; category and comparison probes cite directories → Probe coverage = 1, Citation quality = 1.

• Location pages are thin templates with interchangeable blurbs → Source extractability = 1.

• No probe log and no generative AI report review yet → Measurement loop = 0.

Total: 6/12. The first actions are not “publish 50 AI blogs.” They are unblock OAI-SearchBot, rewrite the top three commercial pages with retrofit decision criteria and local proof points, then re-run the same probe set in 30 days. Only after eligibility and extractability improve does a broader GEO program become the efficient lever.

Decision matrix: what to do with your score

Limitations and common failure patterns

• Probe bias: logged-in personalization, location, and model updates can change answers. Use a consistent setup and treat results as trends, not courtroom proof.

• Surface confusion: Google AI features, ChatGPT search, and other assistants do not share one ranking system. Fix foundations that transfer; do not expect identical citations everywhere.

• Hack chasing: Google explicitly advises against relying on unsupported “AEO/GEO hacks” such as mandatory llms.txt for Google Search AI features or inauthentic mention schemes.

• Click myopia: Google has reported that clicks from pages with AI Overviews can be higher quality in terms of time on site. Judge outcomes by engaged visits and conversions, not click volume alone.

Recommended next steps

1. Complete the six-dimension scorecard this week.

2. Repair any eligibility blockers for Googlebot and OAI-SearchBot.

3. Upgrade the five commercial pages most likely to earn citations on your probe set.

4. Re-run the same probes in 30 days and compare visibility rate and citation rate.

5. If eligibility is clean and category prompts still ignore you, evaluate a focused generative engine optimization program alongside ongoing SEO fundamentals.

If you want a second set of eyes on the audit findings and a prioritized 90-day plan for AI-search visibility, book an Oasbit growth strategy session.

Sources

• Google Search Central: AI features and your website

• Google Search Central: Optimizing for generative AI features on Google Search

• Google Search Central Blog: Succeeding in AI search

• Search Console Help: Generative AI performance report (Search)

• OpenAI Help Center: ChatGPT Search

• OpenAI Help Center: Publishers and Developers FAQ