How to Do an AI Search Audit (with Free Template)
| Most brands are invisible in AI search without knowing it. An AI search audit – covering entity mapping, crawlability, schema, content freshness, off-page citations, competitor prompt analysis, and tracking setup – shows you exactly where you stand and what to fix. This guide walks through every step, with a free downloadable template and an Atlas trial at the end. |
What’s in This Audit
- Why Every Brand Needs an AI Search Audit Right Now
- Step 1: Entity Mapping
- Step 2: Crawlability Check
- Step 3: Schema Audit
- Step 4: Content Freshness Review
- Step 5: Off-Page Citation Audit
- Step 6: Competitor Prompt Analysis
- Step 7: Atlas Tracking Setup
- Industry Updates: The Numbers That Make This Urgent
- How to Score Your Audit Results
- FAQ
Why Every Brand Needs an AI Search Audit Right Now
The Dropbox marketing team ran an audit. What they found was uncomfortable.
Their brand was mis-represented in AI overview summaries. Not missing – actively described in ways that contradicted their strategy. That audit became the catalyst for a board-level shift in how Dropbox approached GEO.
| “The audit was very specific about where we were underrepresented, how we were underrepresented, and our brand was actually mis-represented from a plethora of content out there. That was the unlock – when you start to put it in revenue terms, you really go: I need to pay attention.”– Enterprise marketing leader (Dropbox), Pepper Index event |
Most brands are in this position. They know traffic is softening. They sense something has changed. But they haven’t run the numbers.
Here’s what the data shows. AI search visits grew 42.8% year over year in Q1 2026, while Google grew just 2.4%. Google AI Overviews now reach 2 billion monthly users. And less than 20% of Google’s top-10 results appear in AI-cited sources – meaning you can rank well on Google and be completely invisible in AI answers.
The audit tells you which side of that gap you’re on.
According to Pepper’s own benchmarking across 110 enterprise companies, 69% have no llms.txt file, 61% have a weak entity footprint, and 60% have no tracking of retrievability in AI answers at all. The audit is the diagnosis. Atlas is the ongoing monitoring. This guide gives you both.
| DEFINITION: What is an AI Search Audit? An AI search audit is a structured, repeatable process for evaluating whether your brand can be found, understood, and cited by LLM-powered answer engines including ChatGPT, Perplexity, Gemini, and Google AI Overviews. It covers seven layers: entity recognition, technical crawlability, schema markup, content freshness, off-page citation presence, competitive positioning, and ongoing tracking setup. |
Step 1: Entity Mapping
This is where every AI search audit starts – and where most teams discover their biggest blind spot.
LLMs don’t think in keywords. They think in entities. An entity is any unique, identifiable thing: a brand, a person, a product, a concept, a place. When an LLM encounters your brand, it doesn’t match keywords – it resolves your entity against its knowledge graph. If that entity is incomplete, inconsistent, or missing, the model either gets you wrong or ignores you entirely.
| “AI thinks in entities, not keywords. LLMs actually build knowledge graphs about your brand – asking: what is this thing? What category does it belong to? What relationships connect it to other entities? Add entity checking to your content review process – there are free tools that do this, treat it like Grammarly for AI search.”– Kishan Panpalia, Pepper Index event |
| STEP 1: Entity Mapping | |
| Wikipedia entry | Does your brand have a Wikipedia page? Is it accurate and up to date? |
| Wikidata record | Is your brand listed on Wikidata with correct properties: founded date, HQ, founder, website, industry? |
| Crunchbase profile | Is your Crunchbase entry complete and consistent with Wikipedia? |
| Brand name consistency | Is your brand referred to consistently across all owned pages, PR, G2, and third-party mentions? No mix of old vs. new names? |
| Founder entity | Is your founder (e.g. Anirudh Singla) linked to the company entity on LinkedIn, Wikipedia, and Wikidata? |
| Category classification | Do LLMs correctly understand what category your brand belongs to? Test: ask ChatGPT ‘What does [Brand] do?’ – is the answer accurate? |
| TOOL TIP: Tool tip: Use Google’s Natural Language API (free) to run entity extraction on your homepage and key pages. Check which entities are detected and at what salience score. | |
Step 2: Crawlability Check
Your content can’t get cited if AI crawlers can’t access it. This step checks that you aren’t accidentally blocking the systems you’re trying to be cited by.
One enterprise brand discovered they had been blocking all AI bots in their robots.txt for 8 months – a configuration error that caused a 94% drop in AI visibility before it was caught. This check takes 15 minutes and can uncover critical issues.
| STEP 2: Crawlability Check | |
| robots.txt | Verify GPTBot, PerplexityBot, GoogleBot, BingBot, and ClaudeBot are not blocked. Confirm with: yourdomain.com/robots.txt |
| llms.txt | Does an llms.txt file exist at yourdomain.com/llms.txt? This is the AI-search equivalent of robots.txt – it tells LLM crawlers which pages to prioritise. |
| GPTBot crawl in GSC | In Google Search Console, filter crawl stats by GPTBot. Are key pages being crawled? What is the crawl frequency? |
| Page speed | Do target pages load in under 2 seconds? Slow pages get deprioritised during RAG retrieval. Test with PageSpeed Insights. |
| Canonical tags | Do all key pages have correct canonical tags? Duplicate or self-referencing canonicals confuse AI crawlers. |
| Indexation | Are all target pages returning 200 status codes? Submit sitemap to Bing Webmaster Tools (powers ChatGPT Search and Copilot). |
| TOOL TIP: Tool tip: Google Search Console (free) for GPTBot crawl data. Bing Webmaster Tools (free) for Copilot/ChatGPT search indexation. Both should be set up before running any other audit step. | |
Step 3: Schema Audit
Schema markup is how LLMs understand entity relationships. Without it, even well-written content gets misclassified or ignored.
This step checks whether the right schema types are implemented on the right pages – and whether they’re valid. Invalid schema is nearly as bad as missing schema: the model receives a signal it can’t parse.
| STEP 3: Schema Audit | |
| Organization schema | Is Organization schema implemented on the homepage with: name, url, logo, description, sameAs (Wikipedia, LinkedIn, Crunchbase)? |
| Article schema | Does every blog post have Article schema including author (with Person schema), datePublished, headline, and description? |
| FAQPage schema | Does every page with an FAQ section have FAQPage schema? This directly enables LLM extraction of Q&A pairs. |
| Person schema | Is there a Person schema for named executives linked to the company entity? |
| SoftwareApplication schema | If you have a product (e.g. Atlas), does the product page have SoftwareApplication schema with applicationCategory and description? |
| DefinedTerm schema | For any key concept your brand owns or defines (e.g. Search Everywhere Optimization), is DefinedTerm schema implemented? |
| TOOL TIP: Tool tip: Validate all schema with Google Rich Results Test (free) and Schema Markup Validator. Invalid schema is worse than no schema – it sends a corrupted signal. Use Screaming Frog to crawl and identify schema gaps across the whole site. | |
Step 4: Content Freshness Review
LLMs weight recency. Content that hasn’t been updated signals stale authority.
This step isn’t just about updating publish dates. It’s about identifying which pages are most relevant to the prompts your buyers are running – and making sure those pages are current, well-structured, and contain original, citable data.
| STEP 4: Content Freshness Review | |
| Last updated dates | Identify your 20 highest-traffic pages. When were they last updated? Flag any page not updated in the last 6 months. |
| Content structure | Do target pages use atomic 2-4 sentence blocks, H2/H3 headers as question blocks, and FAQ sections? (See: How to Structure Content for AI Citation) |
| Data freshness | Do pages contain statistics or data points older than 12 months? Outdated stats lower citation credibility – update with current sources. |
| TLDR blocks | Does every key page have a TLDR in the first 30% of content? This directly improves citation probability (44.2% of citations come from the first 30% of a page). |
| Author attribution | Is every piece of content attributed to a named author with visible credentials? Anonymous content has weak E-E-A-T signals. |
| Comparison content | Do you have dedicated comparison or alternative pages targeting ‘[competitor] alternative’ queries? These carry the highest citation weight in B2B categories. |
| TOOL TIP: Tool tip: Use Screaming Frog to export all pages with their last modified date. Filter by pages not updated in 6+ months. Cross-reference with your top 100 target prompts to prioritise which pages to refresh first. | |
Step 5: Off-Page Citation Audit
Your website isn’t where most AI citations come from. This is the step most teams skip – and the one with the biggest impact.
According to Pepper’s GEO research, authoritative list inclusions carry 40-65% citation weight in LLM responses. Review platforms like G2 carry up to 70%. YouTube is the single most-cited domain in LLM responses. Your owned website, by comparison, carries low citation weight on its own.
This step maps where you currently appear – and where you’re invisible.
| STEP 5: Off-Page Citation Audit | |
| G2 / Capterra listing | Is your brand listed on G2? How many reviews? Is your product description current? G2 alone drives 14 LLM citation pages. |
| Authoritative list presence | Are you featured in ‘Best [category] tools’ or ‘Top [category] platforms’ articles on high-DA sites? Run: site:searchengineland.com OR site:g2.com ‘[Brand name]’ in Google. |
| YouTube channel | Does your brand have a YouTube channel with GEO/product content? YouTube is the #1 most-cited domain across LLM responses (95 pages cited). |
| LinkedIn Articles | Has your founder published LinkedIn Articles (not posts – Articles) in the last 6 months? Articles are indexed by Google and cited by Perplexity. |
| Reddit / Quora presence | Is your brand mentioned in relevant subreddits or Quora spaces? Reddit is the 3rd most-cited domain in LLM responses (44 pages). |
| PR / editorial coverage | How many editorial mentions from DA 50+ publications does your brand have in the last 12 months? 61% of LLM responses are influenced by trusted editorial sources. |
| TOOL TIP: Tool tip: Run your brand name in ChatGPT, Perplexity, and Gemini right now. Note which sources they cite when they mention you (or a competitor). Those sources are your highest-priority off-page investment targets. | |
Step 6: Competitor Prompt Analysis
This step tells you exactly who is winning the queries you should be winning – and why.
The Dropbox team described this experience at Pepper’s Index: tracking ‘prompt clusters’ the way they once tracked keyword clusters. Knowing which prompts your competitors own, which sources LLMs use to cite them, and what content structure drives those citations is the most actionable intelligence in a GEO programme.
| STEP 6: Competitor Prompt Analysis | |
| Build a prompt list | Create a list of 20-30 prompts your buyers actually run: category queries (‘best [category] for [use case]’), comparison queries (‘[Your Brand] vs [Competitor]’), and problem queries (‘how to solve [problem your product addresses]’). |
| Run prompts across 3 LLMs | Test each prompt in ChatGPT, Perplexity, and Gemini. Log: which brands appear, in what position, and which URLs are cited as sources. |
| Map competitor citation sources | For each competitor that appears, identify the sources cited. These are the pages you need to appear on or outperform. |
| Identify gaps | Which prompts return zero mentions of your brand? Flag these as priority GEO content targets. |
| Check brand representation accuracy | When your brand does appear, is the description accurate? Does it reflect your current positioning? Inaccurate representation can be worse than absence. |
| Quantify the gap | Count: competitors mentioned vs. you mentioned, across your full prompt list. This is your baseline share-of-answer score. |
| TOOL TIP: Tool tip: Atlas (atlas.pepper.inc) automates this across all major LLMs simultaneously – running your full prompt list, logging citations, tracking competitor positions, and generating a share-of-answer score. For a manual audit, use a Google Sheet with columns: LLM, Prompt, Brand Mentioned, Citation URLs, Position. | |
Step 7: Atlas Tracking Setup
The audit gives you a baseline. Atlas makes it a system.
Running an audit manually once a quarter gives you a snapshot. The brands winning AI search are tracking continuously – monitoring citation shifts weekly, catching competitor moves early, and updating content the moment it drops out of LLM answers.
A Dropbox marketing leader described the outcome at Pepper’s Index: LLM impact is now a line item on their weekly scorecard, alongside pipeline. They track prompt clusters instead of just keywords.
| “Now, even in our weekly scorecard, there’s a line item that shows you the funnel, the LLM impact. We can actually see what’s working, what’s not. We’re talking in terms of prompt clusters now, the way we used to talk about keyword clusters.”– Enterprise marketing leader (Dropbox), Pepper Index event |
| STEP 7: Atlas Tracking Setup | |
| Enter brand name variants | In Atlas, add all versions of your brand name LLMs might use – including old names, abbreviations, and product names. |
| Set up competitor tracking | Add 3-5 direct competitors. Atlas will show you citation share side-by-side so you can see the gap clearly. |
| Load your prompt list | Upload the 20-30 prompts from Step 6. Atlas scans these across ChatGPT, Gemini, Perplexity, and Copilot weekly. |
| Configure alert thresholds | Set an alert if any competitor gains 5+ citations in a single week – early warning for competitive moves. |
| Set up GA4 AI channel | In Google Analytics 4, create a custom channel group capturing sessions from chatgpt.com, perplexity.ai, claude.ai, and gemini.google.com – these often misclassify as ‘direct’ by default. |
| Establish baseline report | Export your Month 1 Atlas report as a PDF. This is your benchmark – every subsequent month is measured against it. |
| TOOL TIP: Start your Atlas trial at atlas.pepper.inc. The free trial includes one full brand audit across all major LLMs, a competitor gap report, and your first share-of-answer score. | |
Industry Updates: The Numbers That Make This Urgent
AI Search Traffic Up 42.8% YoY
AI search visits grew from 15.6 billion to 27.4 billion in Q1 2026 – a 42.8% year-over-year increase, while Google grew just 2.4%. AI search is no longer a trend. It is the primary growth surface for organic discovery (Onely/Ahrefs, 2026).
2 Billion Users See Google AI Overviews
Google AI Overviews reached 2 billion monthly users in 2025, reported by CEO Sundar Pichai. The share of organic keywords triggering an AI Overview grew from 1.5% to 32% in a single year – a 20x increase. AI Overviews reduce CTR by 58% for the top-ranking organic page (Ahrefs, 2025).
85.7% of Businesses Are Invisible in AI Responses
Despite AI search’s explosive growth, 85.7% of businesses remain invisible in AI-generated answers. Only 16% of brands actively track their AI search visibility (Botric, 2026). The audit gap is enormous – and it’s an opportunity for brands that move now.
62% of Consumers Consult AI Before Purchasing
According to a Salesforce survey, 62% of consumers now consult AI assistants before making a purchase decision. In B2B, over 50% of buyers start research in an LLM, up from 24% just 18 months ago (G2 research, presented at Pepper Index). The discovery funnel has shifted – the audit tells you whether you’re in it.
Pepper’s Enterprise Benchmark: 69% Have No llms.txt
Across 110 enterprise companies audited by Pepper, 69% have no llms.txt file, 61% have a weak entity footprint, 70% have unstructured content with no FAQ or bullet formatting, and 60% have no retrievability tracking in AI answers. These aren’t minor gaps – they’re foundational failures that compound over time.
How to Score Your Audit Results
Use this scoring guide after completing all 7 steps. Each completed check item is 1 point. Maximum score: 36 points.
| Score Range | What It Means | Priority Action |
| 0-12 | Critical: you are likely invisible in AI search | Start with Steps 1 (entity) and 2 (crawlability) – foundational fixes unlock everything else |
| 13-20 | Partial visibility: significant gaps in specific layers | Focus on Steps 3 (schema) and 5 (off-page) – these have the highest citation weight impact |
| 21-28 | Competitive: visible but not dominant | Prioritise Steps 6 (competitor prompts) and 7 (Atlas tracking) to find and close specific gaps |
| 29-36 | Strong: well-positioned for AI search leadership | Move to continuous optimisation with Atlas – track weekly, iterate monthly, expand prompt coverage |
| “GEO is not a one-time fix. It’s a monitoring programme. The brands winning AI search aren’t the ones who ran an audit once – they’re the ones tracking prompt clusters every week.”– Anirudh Singla, Founder & CEO, Pepper |
| Get the Free AI Search Audit Template + Atlas Trial Download the 7-point audit template pre-loaded for your team – then connect it to Atlas for automated, ongoing GEO tracking across ChatGPT, Perplexity, and Gemini. → Download template + start Atlas free trial at atlas.pepper.inc |
FAQ
What is an AI search audit?
An AI search audit is a structured process for evaluating whether your brand can be found, understood, and cited by AI-powered answer engines like ChatGPT, Perplexity, and Gemini. It typically covers entity recognition, technical crawlability, schema markup, content structure, off-page citation presence, competitor positioning, and ongoing tracking setup.
How often should you run an AI search audit?
Run a full 7-step audit quarterly, with lighter citation footprint checks monthly. AI engines update citation behaviour frequently, competitors publish continuously, and content freshness decays. A quarterly cadence with logged baselines lets you measure whether fixes actually improved citation frequency.
How long does an AI search audit take?
A manual audit using this 7-step framework takes approximately 4-6 hours for a single domain. An automated audit using Atlas (atlas.pepper.inc) compresses steps 5, 6, and 7 significantly – the full prompt scan runs in minutes across all major LLMs.
What is entity mapping in an AI search audit?
Entity mapping is the process of verifying that your brand, products, and key people are correctly identified, categorised, and connected in LLM knowledge graphs. It checks Wikipedia, Wikidata, Crunchbase, and your own content for consistency. Inconsistent entity data causes LLMs to misrepresent or ignore your brand.
What tools do I need to run an AI search audit?
For a free manual audit: Google Search Console, Bing Webmaster Tools, Google Rich Results Test, Schema Markup Validator, Screaming Frog (free tier), and direct access to ChatGPT, Perplexity, and Gemini. For automated, ongoing auditing: Atlas by Pepper (atlas.pepper.inc) handles steps 5-7 automatically and tracks changes weekly.
Latest Blogs
LLMs don’t rank brands – they recognize them as entities. If your brand isn’t registered in Google’s Knowledge Graph, backed by an Organization schema, anchored on Wikipedia, and consistently named across the web, you’re invisible in AI search. This guide walks you through the complete entity optimization workflow, with a brand audit checklist at the […]
Your writing isn’t the problem. Your structure is. Here’s how to rebuild it for the machines that now decide who gets cited. LLMs don’t read your content like humans do. They extract structured facts. If your content isn’t built for extraction, it won’t be cited.70% of enterprise brands publish unstructured content with no bullets, stats, […]
Most brands are invisible in AI search without knowing it. An AI search audit – covering entity mapping, crawlability, schema, content freshness, off-page citations, competitor prompt analysis, and tracking setup – shows you exactly where you stand and what to fix. This guide walks through every step, with a free downloadable template and an Atlas […]
Get your hands on the latest news!
Similar Posts
Artificial Intelligence
10 mins read
Entity Optimization: How to Make LLMs Recognize Your Brand
Artificial Intelligence
14 mins read
How to Structure Content for AI Citation
Artificial Intelligence
11 mins read