The AI Presence Snapshot: Establishing Your Baseline Before Any Work Begins

No AEO program should begin without a documented baseline. Not a general impression of what AI says about a client. A structured record of what each major AI platform returns for each query category, captured on a specific date, with enough detail that the same exercise can be repeated three months later and the comparison will be legible. Without that baseline, progress in an AEO engagement is a feeling, not a finding.

The baseline serves two functions. It tells you what you are starting from, which determines where effort goes. And it gives you the comparison point that makes future progress demonstrable. A client who asks whether the AEO work is having any effect deserves an answer grounded in before-and-after data, not in platform-specific impressions that cannot be compared across time.

This article covers the four query types every baseline must include, why trust queries are the most consequential category for reputation management clients, how to run the baseline across ChatGPT, Perplexity, and Google AI Mode, and how to structure the documentation so it serves as an actual tracking document.

The Four Query Types Every Baseline Must Cover

A complete AI presence snapshot runs four categories of queries. Each reveals a different dimension of the client's AI reputation and surfaces different types of gaps. Running only brand queries, or only category queries, produces a partial picture that will miss significant problems.

Brand Presence Queries

Brand presence queries test basic entity recognition: does the AI know this brand exists, and if so, what does it say? The simplest form is "Tell me about [Brand Name]" or "What does [Brand Name] do?" These queries reveal whether the AI has indexed sufficient information to generate a confident response, whether the factual attributes it returns are accurate, and what narrative frame it uses to describe the brand.

Inaccuracies at this level are common and often more damaging than practitioners expect. Founding year errors, wrong service area descriptions, outdated leadership names, and category misidentification all appear regularly in AI responses about established businesses. These are not edge cases. The GrackerAI benchmark of 100 cybersecurity vendors found that 73% received zero citations from ChatGPT in their own category, and for those who did appear, factual accuracy varied significantly. Documenting exactly what the AI says at baseline, including any inaccuracies, is the first step toward correcting them.

Category and Service Queries

Category queries test whether the client appears when someone searches the category without naming them: "Best [service type] in [city]" or "Who provides [service] for [use case]?" These are the commercially significant queries where a prospective customer is looking for a provider without a brand in mind. The AI's answer to these queries determines whether the client is visible in the discovery process at all. (Source: Yext AI Citation Behavior Analysis, Q4 2025-Q1 2026.)

The baseline should document not only whether the client appears but also where they appear relative to competitors, which sources the AI cites to support the category response, and whether any competitors are consistently recommended ahead of the client. This data drives the competitive mapping phase.

Trust and Vetting Queries

Trust queries are the most important category for reputation management clients. They are the queries in which the AI's answer most directly determines whether a prospective customer proceeds or walks away: "Is [Brand Name] trustworthy?" "What should I know before hiring [Brand Name]?" "What are the problems with [Brand Name]?" "Has [Brand Name] had any complaints?"

These queries expose the reputation narrative the AI has assembled from its training data and retrieval layer. The answer may draw on aggregate scores from review platforms, Reddit threads, news coverage, regulatory filings, or the client's own content. Whatever sources the AI cites when answering a trust query are the sources forming the client's AI reputation. Identifying those sources at baseline is not optional for an ORM engagement. Listings represent 54.53% of distinct AI citation URLs, but the specific sources driving trust query responses vary significantly by category and geography. (Source: Yext Research, Q1 2026.)

Trust queries should be run for the client and for each named competitor. The comparison reveals whether the client faces a specific trust narrative problem or whether their category generally receives cautionary AI responses, and it identifies whether competitors who appear positively have sources the client lacks.

Competitor Comparison Queries

"[Brand Name] vs. [Competitor]" queries reveal how AI systems frame the competitive relationship. Is the client positioned as the stronger option, the comparable alternative, or the less-recommended choice? Which sources does the AI draw on when making comparative claims? Are the factual claims in the comparison accurate?

These queries also surface a specific vulnerability: AI systems frequently make comparative claims with apparent confidence that are based on outdated or inaccurate source material. A client who was outperformed by a competitor three years ago may still be framed unfavorably in AI comparison responses if the sources the AI relies on predate any improvements in their service or standing. Documenting this at baseline is the first step toward surfacing corrective content.

WHY TRUST QUERIES ARE THE PRIORITY

In traditional SEO-driven ORM, a negative result is visible alongside other results. The user can weigh it against positive content, check multiple sources, and make a judgment. A bad result at position three is serious but not definitive. In AI search, the trust query answer is delivered as a conclusion. A user who asks ChatGPT "is [Brand Name] trustworthy?" receives a synthesized response that draws on whatever the AI has indexed and determined to be the answer. The user is not being offered alternatives. They are receiving the AI's assessment. This is why the AI's answer to trust queries is often the single most consequential thing to track and improve in a reputation management AEO engagement. Document it first. Return to it every reporting cycle.

Which Platforms to Run and Why All Three Matter

The baseline should be run across ChatGPT, Perplexity, and Google AI Mode. Running only one platform produces a misleading picture. Only 11% of domains are cited by both ChatGPT and Perplexity. A client who appears favorably in Google AI Mode but not in ChatGPT has a gap that a Google-only baseline would not reveal. (Source: Profound, cited in Leapd April 2026.)

The three platforms also reflect different user populations. ChatGPT and Claude are the dominant AI referral sources for B2B queries. Perplexity is widely used by analysts, journalists, and researchers, making it particularly significant for professional services clients and for trust queries about brands that appear in media coverage. Google AI Mode captures the largest general consumer audience. For most reputation management clients, all three matter, and all three need to be baselined separately.

Platform-Specific Characteristics to Note at Baseline

ChatGPT: Retrieves through Bing's index. Favors consensus sources, named authors, and listicle formats. Cites 7 to 8 sources per response on average, but only cites approximately 15% of the pages it retrieves. Named authors carry a citation odds ratio of 1.40 versus 1.12 overall: a bylined article is 25% more likely to be cited than anonymous content. (Source: Authority Tech AI Citation Trust Signals, 2026.)

Perplexity: Runs its own index. Weighs freshness at approximately 40% of its ranking signal: content older than 30 days is deprioritized. Cites on 100% of queries, the highest per-query citation rate of any engine at 13.8 per response on average. Reddit accounts for roughly 46.7% of Perplexity's top citations. 80% of Perplexity-cited content does not rank in Google's top search results. (Source: Authority Tech AI Citation Trust Signals, 2026.)

Google AI Mode: Correlates more closely with traditional search ranking signals than the other platforms. AI Overview citations from the organic top 10 have declined from 76% in mid-2025 to approximately 38% by early 2026, but the correlation still exists and is higher here than on ChatGPT or Perplexity. (Source: Ahrefs data, cited in Leapd April 2026.)

Documenting the Baseline for Progress Tracking

The baseline document needs to be structured for comparison, not just for reference. A narrative summary of what AI says about the client is useful context. It is not a tracking document. The tracking document needs fields that can be populated on the same schedule in future reporting cycles so that changes are visible and attributable.

What to Record for Each Query

Date and platform: Essential. AI responses change. A response captured in January may differ significantly by April. Always record when and where.
Full response or detailed summary: The complete AI response, or a structured summary noting every factual claim about the client, every entity mentioned, and the overall sentiment framing.
Sources cited: Every URL or source name that the platform links to or attributes. These are the sources forming the client's AI reputation for this query. Record source type alongside source name using the category codes from the competitive landscape mapping methodology.
Client mentioned: Yes, No, or Partial (mentioned in passing but not recommended or discussed substantively).
Client characterization: Recommended, Neutral, Cautionary, or Negative. This is the sentiment framing of the AI's treatment of the client in this specific response.
Factual accuracy: Any inaccuracies are noted, with the accurate information documented alongside. These are correction priorities.

Reporting Cadence

The baseline is a snapshot, not a monitoring system. After the initial baseline, the same query set should be run monthly for active engagements and quarterly for maintenance engagements. The comparison between the baseline and each subsequent run is where progress becomes demonstrable. A trust query that returned cautionary language at baseline and now returns neutral language represents measurable, reportable progress, even before any improvement in leads or conversions is attributable.

Citation volatility requires frequent checks: Profound research documents 40-60% citation-source drift across major platforms month to month. A baseline run once and never revisited will be substantially out of date within 90 days. Build the re-run cadence into the engagement structure from the start, not as an afterthought when a client asks for evidence of progress.