Brand Identification — How to Map Your Brand Before Running an AI Visibility Audit
Brand identification is the first step of an AI visibility audit — and the one most people underestimate. It feels like setup work: tell the system what your brand is called, list your competitors, move on. In practice, it's the step that determines the accuracy of everything downstream.
AI models don't always refer to your brand the way you refer to yourself. They may use a product name instead of a company name, a common abbreviation, a parent brand, or a variant picked up from third-party content. If you haven't mapped your full brand surface before running an audit, you'll miss appearances, misattribute results, and draw conclusions from an incomplete picture.
This post walks through how to do brand identification properly — what pages to select, how to use an LLM to extract a structured brand profile, how to build your competitive set, and how to interpret the gaps your pages will inevitably leave.
Select Your Pages
The brand identification step is driven by your website content. You're not writing your brand profile manually — you're feeding the right pages to an LLM and letting it extract a structured profile from what's actually published about you.
What to include:
- Your homepage — your primary framing of what you do and who you serve
- Core product or service pages that describe your main offering in detail
- Your about page, for parent brand relationships and company-level positioning
- Specific capability or use-case pages if your product spans multiple areas
When you provide pages, tell the LLM which page is your primary source versus supporting material. This helps it weight positioning signals correctly — your homepage carries more canonical brand weight than a product sub-page.
Why it matters: Page selection quality determines profile quality. Homepage-dominated page sets produce profiles that are strong on broad positioning language but thin on capability depth, differentiation specifics, and audience nuance. The richer and more representative your page set, the more accurate the brand profile that comes out of it — and the more accurate everything the audit produces downstream.
Extract Your Brand Profile
Once you have your pages, use an LLM to extract a structured brand profile. The output should go beyond a simple description of what your brand does. A useful brand profile should cover:
- What the brand does and who it serves — a clear, grounded overview in plain language
- The problem it exists to solve — in the language the brand itself uses, not a generic restatement
- The core offering — what the product actually is and how it delivers on the problem
- Key capabilities — the specific things it can do, as a concrete list
- Positioning language — the specific phrases, taglines, and framing the brand uses verbatim, not summaries of them
- Differentiators — what the brand claims makes it distinct from alternatives
- Market context — the broader shift or trend the brand is positioning itself within
Why it matters: AI models form their understanding of your brand from what's published about you — on your site and elsewhere. A structured brand profile gives you a clear, auditable picture of what that content actually says, in the terms an AI model would use to describe you. It also surfaces gaps: positioning claims that aren't evidenced anywhere, capabilities mentioned only in passing, audience framing that's implied but never made explicit. Knowing these gaps is the first step to closing them.
Identify Your Audience Segments
As a byproduct of the brand profile extraction, ask the LLM to identify the audience segments implied by your content — the distinct buyer types your pages appear to be written for.
For each segment, you want to understand:
- Who they are — their role, seniority, and the organizational context they're operating in
- Their role in the decision — are they the buyer, the evaluator, an end user, or an implementer? Each role asks different questions and weighs different criteria
- Their likely goals — what they're trying to achieve when they turn to an AI assistant for help
- Their likely pain points — the problems driving their search
- Their likely AI question types — whether they'd ask diagnostic questions (figuring out a problem), educational questions (learning how something works), comparison questions (evaluating options), or transactional questions (ready to select a vendor)
Not every audience segment your pages imply will be worth testing. Some will have weak evidence — they're inferred from a single line of copy rather than a consistent thread of messaging. Some will be overlapping in terms of goals, pain points and likely question types. Objective in this step should be to find as many audience segements as possible. You will create concrete user personas from these audience segments in the next step.
Why it matters: This is the direct input to Step 2 — persona design doesn't start from scratch, it starts from here. Surfacing audience segments from your brand content also tells you something important: the buyers your pages speak to clearly, and the buyers your content has been ignoring. Both are useful signals before a single test conversation runs.
Build Your Competitive Set
The competitive set tells the audit which brands to track alongside yours in every conversation. AI models frequently cite, mention or recommend competitors when they don't surface your brand — and you need to know which competitors are winning those conversations and why.
For each competitor, capture:
- Primary name — the canonical name they use for themselves
- Website — their primary domain
- Aliases — other names they're commonly referred to by, including shortened forms, product names under a parent brand, and category-specific labels. A brand like "HubSpot Service Hub" might appear as "HubSpot" or "Service Hub" in AI responses. If you only track the full name, you'll miss a significant portion of appearances.
- Alternate domains — secondary URLs associated with the brand, since citation sources may vary between a main domain and a product-specific subdomain
How to build the list: Start with who you lose deals to, not who appears on your category page. Ask your sales team who prospects mention before they've narrowed to your specific category. Establish a list of 6–10 brands — enough to reveal meaningful displacement patterns without diluting the analysis. You can also ask an LLM to identify your competitors.
Why it matters: Competitor displacement is one of the most actionable findings an AI visibility audit produces. But it's only actionable if you're tracking the right competitors, under the names AI models actually use to refer to them. A competitor whose aliases aren't mapped will appear as an uncredited citation, and a displacement pattern you can't attribute is a pattern you can't respond to.
Review What Your Pages Couldn't Tell the Model
A well-prompted brand profile extraction should include an honest assessment of its own limitations — what the provided pages didn't give the LLM enough to work with.
Common limitations to look for:
Thin differentiation evidence. If your pages describe what you do but don't explicitly contrast it with alternatives, your differentiators will be inferred from positioning language alone. This is softer than evidenced comparison content, and AI models tend to surface the latter in evaluation-intent queries.
Missing pricing and packaging detail. Homepage-level content rarely includes packaging specifics. If AI-generated answers about your brand regularly describe your pricing or target segment inaccurately, the root cause is often that this information isn't well represented in indexed content.
Parent brand ambiguity. If your product sits under a larger company umbrella — and the parent brand also has a strong web presence — AI models may conflate the two. The brand profile should explicitly distinguish product from company, and map which aliases belong to which.
Narrow audience evidence. If your pages are primarily written for one buyer type, the audience segments your profile surfaces will skew toward that audience. The full range of stakeholders in your buying process may not appear.
Why it matters: These limitations aren't failures of the extraction process — they're accurate signals about where your web content doesn't yet give AI models enough to work with. That's exactly what an AI visibility audit exists to find. The limitations section of your brand profile is often the clearest early indicator of which content gaps will show up as visibility problems when the conversations run.
What You Have at the End of This Step
A completed brand identification step gives you three things the rest of the audit depends on:
A defined brand surface — the canonical name, aliases, and associated domains the audit will use to recognize appearances of your brand across AI-generated conversations.
A structured brand profile — the positioning, capabilities, and problem framing the audit will use to evaluate whether AI responses accurately describe your brand, partially describe it, or mischaracterize it.
A preliminary audience map — the buyer segments implied by your content, each with goals, pain points, and likely AI question types. This is the input to Step 2.
A competitive set — the list of competitors with their names, aliases and websites. This will be used in Step 6 for identifying your brand displacements.
What a Brand Profile Actually Looks Like
Below is a trimmed example from a real brand identification run on Freshdesk, generated from five pages: the Freshdesk Omni homepage, the core Freshdesk product page, the live chat software page, the Freshworks homepage, and the Freshworks about page.
**Brand:** Freshdesk (parent: Freshworks)
**Primary offering:** Freshdesk Omni
**Aliases:** Freshdesk Omni, Freshworks Freshdesk, Freshdesk by Freshworks
**Primary category:** Customer service software
**Secondary categories:** Omnichannel support software, Help desk software, Ticketing software, Live chat and chatbot software
---
**What Freshdesk does**
An AI-powered customer service platform that unifies conversations, tickets, self-service, workflows, and analytics in one workspace. Teams can resolve issues faster using a combination of omnichannel support, advanced ticketing, and AI features including AI Agents, AI Copilot, and AI Insights.
**Problem it solves**
Customer service teams working across disconnected channels, clunky workflows, and separate tools that slow resolution and reduce context. The platform also addresses pressure to automate repetitive requests and improve agent productivity without adding implementation complexity.
**Positioning language** *(verbatim phrases from the pages)*
Customer service that puts people first · People-first AI · Enterprise capability without enterprise complexity · Uncomplicating customer service · Free to try. Fast to scale. · Resolution ready
**Key differentiators** *(as claimed on the pages)*
Unified workspace combining conversations, context, and AI — Conversational support and ticketing in one offering — Ease of setup and scaling — Positioned against bloated, hard-to-implement legacy software
---
**Audience segments identified: 14 total — 6 with strong evidence, 6 moderate, 2 weak**
*Two example segments:*
**Customer service leaders** — *Buyer · High commercial relevance · Strong evidence*
Leaders responsible for support performance who want unified visibility, better resolution outcomes, and scalable AI-assisted delivery.
- Goals: Improve resolution rates, unify support channels and reporting, scale efficiently, improve customer experience
- Pain points: Fragmented tools, limited cross-channel visibility, slow resolutions, pressure to scale without adding headcount
- Likely AI questions: *"What are the best AI-powered customer service platforms for omnichannel support?"* · *"Which customer service tools offer AI insights for support leaders?"*
**Support operations admins** — *Implementer · High commercial relevance · Strong evidence*
Admins configuring channels, automations, routing, integrations, and business rules for support systems.
- Goals: Enable new channels quickly, configure automations and routing, reduce implementation complexity
- Pain points: Complex setup across channels, integration friction, slow implementation, hard-to-maintain workflows
- Likely AI questions: *"Which customer service platforms are easiest to configure for omnichannel support?"* · *"What help desk tools connect easily with business apps?"*
---
**Competitive set (8 competitors)**
- Zendesk *(aliases: Zendesk Suite, Zendesk Support, Zendesk for Service)*
- Intercom · Help Scout *(alias: Helpscout)* · Zoho Desk *(alias: Zoho)*
- Salesforce Service Cloud *(aliases: Salesforce, Service Cloud)*
- Front
- LiveAgent *(alias: Live Agent)*
- Gorgias
- Kustomer
- HubSpot Service Hub *(aliases: HubSpot, Service Hub)*
---
**Profile limitations flagged**
- Page set is dominated by homepage-level marketing content — limited pricing, packaging, and comparison detail
- Differentiators are expressed through positioning language rather than explicit head-to-head comparisons — softer evidence than dedicated comparison pages would provide
- Some ambiguity between Freshdesk (product brand), Freshdesk Omni (primary offering), and Freshworks (parent company) — not fully resolved by the supplied pages
- Market context conclusions rely partly on broad positioning language rather than explicit category analysis
---
These limitations aren't gaps in the extraction — they're accurate signals about where the web content doesn't yet give AI models enough to work with. Closing them is the work the rest of the audit will prioritize.
Next in this series: Step 2 — Persona Design: how to turn the audience segments from your brand profile into test-ready personas that reflect how real buyers actually interact with AI assistants.
If you'd rather see what your brand's profile looks like before doing this yourself, request a diagnostic run.
By Gaurav