Setting Up an AI Brand Baseline: First Steps

Why You Need an AI Brand Baseline

Every optimization effort begins with measurement. In traditional SEO, you check your Google rankings before you start optimizing. In paid advertising, you record your baseline cost-per-click and conversion rate. AI SEO is no different -- except that most companies skip this step entirely, and then struggle to demonstrate progress three months later.

An AI brand baseline answers a simple question: how does AI see your brand right now? Not how you think it sees you. Not how you hope it sees you. How it actually, demonstrably sees you today. The answer is often surprising. A company that dominates Google search results may discover that ChatGPT does not mention them at all. A business with excellent customer reviews may find that Gemini associates them with a competitor. A well-known brand may learn that Perplexity cites outdated information from three years ago.

The baseline creates accountability. When you invest in AI SEO -- whether through internal resources or an agency -- the baseline is the "before" photo. Without it, the conversation three months later becomes "I think we're doing better" instead of "We went from being mentioned in 15% of relevant AI queries to 47%."

Research shows that 88% of pages cited by AI models are not in Google's top 10, which means your Google rankings tell you almost nothing about your AI visibility. Your current search console data, no matter how detailed, does not capture how AI perceives your brand. You need a separate measurement.

For a foundational understanding of why AI visibility matters, see our guide on what AI SEO is.

The Five Dimensions of an AI Baseline

A thorough baseline captures more than just "does AI mention us?" It measures five distinct dimensions that together paint a complete picture of your AI brand presence:

1. Presence

Is your brand mentioned at all? For each query, record a simple yes or no. This is the most fundamental metric. If AI does not mention you, nothing else matters.

2. Accuracy

When AI does mention your brand, is the information correct? Check company descriptions, product details, pricing, location, founding date, team members, and any specific claims. Inaccurate information is worse than no mention -- it actively misleads potential customers.

3. Sentiment

What tone does the AI use when discussing your brand? Is it positive, neutral, or negative? Does the AI recommend you, present you as one option among many, or warn users about potential issues? Sentiment directly impacts whether an AI mention converts into a customer.

4. Citation

Does the AI link to your website? There is a critical difference between being mentioned by name and having a clickable citation. A mention without a link gives you brand awareness but no traffic. A citation with a link drives visitors directly to your site -- and AI referral traffic converts at 4.4x the rate of organic search.

5. Positioning

Where in the response does your brand appear? First mention? Last mention? In a list of five alternatives? AI responses have a structure, and appearing first or as the primary recommendation carries more weight than being the fourth option in a bullet list.

Building Your Query List

The quality of your baseline depends entirely on the quality of your query list. Poor queries produce a misleading baseline. Here is how to build a comprehensive, representative set.

Category 1: Brand Queries (5-8 queries)

These test whether AI knows who you are. Examples:

• "What is [your company name]?"
• "Tell me about [your company name]"
• "[Your company name] reviews"
• "Is [your company name] reliable?"
• "Who founded [your company name]?"
• "[Your company name] vs [competitor]"

Category 2: Product/Service Queries (5-8 queries)

These test whether AI associates your products or services with your brand:

• "Best [your product category] for [target audience]"
• "What are the top [your service] providers?"
• "[Your product type] recommendations"
• "How much does [your service] cost?"

Category 3: Industry/Topical Queries (5-8 queries)

These test whether AI considers you an authority in your space:

• "[Your industry] trends 2026"
• "How to [problem your product solves]?"
• "What should I look for in a [your product category]?"
• "[Industry-specific question your content answers]"

Category 4: Local Queries (3-5 queries, if applicable)

If your business has a local component:

• "Best [your service] in [your city]"
• "[Your product category] near [your location]"
• "[Your industry] companies in [your region]"

Category 5: Competitor Comparison Queries (3-5 queries)

These reveal how AI positions you relative to competitors:

• "[Your company] vs [competitor A]"
• "Best alternatives to [competitor A]"
• "Compare [your product] and [competitor product]"

Aim for 20-30 queries total across these categories. This gives you enough data points for a meaningful baseline without making the process unmanageably long.

To check your current status before building the full baseline, start with our quick guide on checking if your site is visible in AI.

Which AI Platforms to Test

Not every AI platform is equally important, but a comprehensive baseline should cover at least the top five. Here is the priority order and what makes each platform unique:

| Platform | Priority | Why Include It | |---|---|---| | ChatGPT | Essential | 84.2% of AI referral traffic; largest user base | | Google Gemini | Essential | Integrated into Google Search; 75M+ active users | | Perplexity | Essential | Fastest-growing AI search; strong citation behavior | | Claude | High | Apple Safari integration; growing enterprise adoption | | Microsoft Copilot | High | Built into Windows, Office, and Bing | | Grok | Optional | Relevant if your audience is active on X/Twitter | | DeepSeek | Optional | Relevant for technical/developer audiences |

Important: Use the free tier or standard consumer version of each platform for your baseline queries. Enterprise or API versions may produce different results. You want to see what your potential customers see.

Account and Settings Considerations

Log out of any accounts before querying, or use incognito/private browsing mode. Personalized AI responses based on conversation history or account preferences can skew your baseline. You want the default, unpersonalized response that a new user would receive.

For ChatGPT, test with both GPT-4o and any available web-browsing modes. For Gemini, test both the standalone Gemini app and Google Search AI Mode if available. These can produce different results.

How to Run the Manual Baseline

With your query list and platform list ready, here is the step-by-step process for running your manual baseline. Set aside 2-4 hours for a thorough 25-query, 5-platform baseline.

Step 1: Prepare Your Environment

Open each AI platform in separate browser tabs or windows. Use incognito mode. Have your tracking spreadsheet open and ready for data entry. Set a timer -- you want to complete this in a single session to ensure consistent results.

Step 2: Run Queries Systematically

Work through your query list one query at a time across all platforms. This means: enter Query 1 into ChatGPT, record the result, enter Query 1 into Gemini, record the result, and so on across all platforms before moving to Query 2.

This approach is better than running all queries on one platform first because it keeps your recording consistent and prevents the AI from building context from previous queries in the same session.

Step 3: Record Results Immediately

For each query-platform combination, record:

• Date and time of the query
• Exact query text (word for word)
• Platform and version (e.g., "ChatGPT GPT-4o" or "Gemini 2.0")
• Was your brand mentioned? (Yes/No)
• What was said? (Copy the relevant excerpt)
• Was information accurate? (Yes/No/Partially, with notes)
• Sentiment (Positive/Neutral/Negative)
• Was your website cited/linked? (Yes/No)
• Position in response (1st mention, 2nd, in a list, etc.)
• Screenshot (take one for every query -- responses change over time)

Step 4: Handle Edge Cases

Sometimes AI responses are ambiguous. Your brand might be mentioned indirectly ("companies like yours" without naming you), or mentioned in a way that could apply to multiple brands. Record these as "partial mentions" and note the ambiguity. Do not inflate your baseline by counting vague references as clear mentions.

If an AI platform asks a clarifying question instead of answering directly, record the clarifying question and then provide a reasonable follow-up. Document both the initial response and the follow-up.

Creating Your Tracking Spreadsheet

A structured spreadsheet is the backbone of your baseline. Here is the recommended structure:

Sheet 1: Raw Data

| Column | Content | Example | |---|---|---| | A: Date | Query date | 2026-03-22 | | B: Query | Exact query text | "best CRM for startups" | | C: Category | Query category | Product/Service | | D: Platform | AI platform | ChatGPT GPT-4o | | E: Mentioned | Yes/No | Yes | | F: Excerpt | Relevant text from response | "For startups, [Brand] offers..." | | G: Accurate | Yes/No/Partial | Partial | | H: Accuracy Notes | Details on inaccuracies | "Pricing listed as $49, actual is $39" | | I: Sentiment | Positive/Neutral/Negative | Positive | | J: Cited | Yes/No | Yes | | K: Citation URL | Link if provided | https://yourdomain.com/pricing | | L: Position | Location in response | 1st of 3 recommended | | M: Screenshot | Link to screenshot file | baseline_001.png |

Sheet 2: Platform Summary

Create a pivot table or summary that shows performance per platform:

• Total queries tested per platform
• Mention rate (% of queries where brand appeared)
• Accuracy rate (% of mentions that were accurate)
• Citation rate (% of mentions that included a link)
• Average sentiment score

Sheet 3: Query Category Summary

Another pivot showing performance by query category (brand, product, industry, local, competitor). This reveals where your AI visibility is strongest and weakest.

Sheet 4: Trend Tracking

Leave space for future baseline runs. Each re-run gets its own date column, allowing you to track changes over time. This is where the real value of a baseline becomes apparent -- you can show month-over-month improvement in every dimension.

Scoring and Interpreting Your Results

Raw data needs interpretation. Here is a simple scoring framework to convert your baseline data into actionable metrics.

Overall AI Visibility Score

Calculate the percentage of all query-platform combinations where your brand was mentioned. This is your top-level metric.

• 0-15% mention rate: Invisible. AI platforms do not know your brand. This is the most common starting point and represents the greatest opportunity.
• 16-35% mention rate: Emerging. AI has some awareness but it is inconsistent. You likely appear for brand queries but not for product or industry queries.
• 36-60% mention rate: Visible. AI recognizes your brand across multiple query types. Focus shifts to accuracy, sentiment, and citation rates.
• 61-85% mention rate: Strong. Your brand is well-represented in AI. Optimize for positioning and ensure accuracy across all platforms.
• 86-100% mention rate: Dominant. You are the go-to source for AI in your space. Maintain and defend this position.

Per-Platform Scores

Calculate the same mention rate for each platform individually. It is common to see significant variation -- you might have 50% visibility on ChatGPT but only 10% on Perplexity. This tells you which platforms need the most attention.

Accuracy and Sentiment Scores

Among the queries where you were mentioned, what percentage had accurate information? What percentage had positive sentiment? These secondary metrics tell you whether AI mentions are helping or hurting your brand.

For a deeper understanding of Share of Voice as a metric, see our guide on AI Share of Voice.

From Baseline to Action Plan

Your completed baseline is not just a report -- it is a strategic roadmap. Each finding points to specific optimization actions.

If Mention Rate Is Low (Under 35%)

Your primary issue is AI discovery. Focus on:

• Ensuring AI crawlers can access your website (check robots.txt)
• Adding Schema markup so AI understands your content structure
• Publishing content that directly answers the queries where you were absent
• Strengthening third-party signals (Wikipedia, industry directories, Reddit mentions)

If Accuracy Is Low (Under 70%)

AI has outdated or incorrect information about you. Focus on:

• Updating your website with clear, current information in structured formats
• Correcting your Google Business Profile, Wikipedia entry, and Wikidata records
• Publishing authoritative content that directly states the correct facts
• Using Organization Schema with accurate details

If Citation Rate Is Low (Mentioned But Not Linked)

AI knows about you but is not sending traffic. Focus on:

• Creating citable content with clear, quotable 50-150 word chunks
• Adding FAQ sections that directly answer common queries
• Ensuring your content provides unique value that AI cannot find elsewhere

If Sentiment Is Negative

AI is mentioning you but not favorably. Focus on:

• Identifying the sources of negative information and addressing them
• Building positive signals through customer reviews, case studies, and expert endorsements
• Publishing content that counters specific negative claims with data and evidence

For a complete AI monitoring framework that builds on your baseline, see our AI visibility monitoring guide.

Setting Your Re-Run Schedule

Mark your calendar:

• Month 1-3 (active optimization): Re-run the full baseline monthly. This is your period of most active change and you need frequent measurement.
• Month 4+: Switch to quarterly re-runs. AI model updates happen roughly every 2-4 months, so quarterly measurement aligns with the pace of change.
• After major changes: Always re-run the baseline after significant website updates, new product launches, or major content publications.