How to measure AI visibility for a HubSpot website https://superschema.ai/hubspot-aeo/measuring-ai-visibility Measure AI visibility with several separate signals: whether systems can access your content, how your business appears in sampled answers, whether the information is accurate, and whether identifiable visitors become qualified customers. A single score hides too much to guide useful decisions. For a HubSpot website, start with a repeatable answer sample and your existing analytics and CRM measurement. Add verified crawler records if your infrastructure provides them. Define the limits of each source before comparing results. If you are still choosing what to improve, our HubSpot AEO guide (https://superschema.ai/hubspot-aeo/guide) connects the measurement plan to page content, access, and structured data. Define the outcome before collecting numbers A company mention and a service recommendation can appear in the same answer, but they are different outcomes. So are a citation and a website visit. Signal: Operational definition: What it establishes Crawl or fetch: A recorded request by a verified agent: Observed access activity Citation: An answer links to or explicitly attributes your page as a source: Source use in that captured answer Mention: Your business is named: Presence in that captured answer Recommendation: Your business is presented as a suitable option for the stated need: Suggested fit in that captured answer Accuracy: Answer facts agree with approved, current information: Reliability of the sampled description Referral: An identifiable visit from an AI service: A recorded arrival on your website Conversion: A visitor completes your defined business outcome: A measurable action, subject to attribution limits Use these definitions consistently. A business named in a warning is a mention, but it should not be coded as a positive recommendation. A link in a source panel can be a citation without the answer recommending the business. Crawl activity, citations, recommendations and conversions require separate evidence. Observing one does not establish the next. Crawl, citation, recommendation and conversion need separate evidence. Do not infer one from another. Build a buyer-question sample Choose questions that reflect real evaluation work. Use sales conversations, customer questions, website search data where available, and the services you actually offer. Include branded questions and questions that do not name your business. A fictional HubSpot consultancy might test: - “What should a B2B team check before migrating its website to HubSpot?” - “Which consultancies help UK B2B companies migrate to HubSpot?” - “Does Cedar Bridge Consulting provide ongoing content operations?” - “What should I ask Cedar Bridge Consulting before starting a migration?” The business is fictional; these prompts illustrate categories, not demonstrated outcomes. Maintain a stable core set. Add new questions as services or buyer concerns change, but keep them separate from the historical set. Otherwise a change in the questions can look like a change in visibility. Record the platform, available model or mode, exact prompt, date, language, relevant location, search setting, and session context. Start fresh sessions where practical, and avoid telling the system to recommend your company unless that is the behavior you intentionally want to test. Repeat samples and preserve the evidence One answer is a snapshot. Repeat prompts across scheduled observation periods and retain the answers, linked URLs, and factual findings. Use the same collection procedure so later comparisons remain interpretable. Report a numerator and denominator. For example, “citations appeared in X of Y captured responses in this question set” is clearer than an unexplained visibility score. Always identify the sample size, collection period, and platforms. A measured rate in your chosen prompts is not market share across all user questions. Code each response for mention, citation, recommendation, and accuracy independently. Mark unclear cases for review rather than forcing a favorable classification. Keep broken citations and links to unrelated pages visible in the findings. Check accuracy against a small approved facts sheet: company name, services, service area, important exclusions, and public commercial terms. Record unsupported claims separately from incorrect ones. An answer may state a price you have never published without contradicting a visible number. Use Google's current reports with their limits Google now documents a Generative AI performance report for Search (https://support.google.com/webmasters/answer/16984139?hl=en). It covers impressions in AI Overviews and AI Mode, with page, country, date, and device dimensions. Its help page says the insights rolled out worldwide on August 31, 2026, while also retaining access and insufficient-impression caveats. Check the report available for your property rather than assuming every account displays identical data. The documented report is an impression view. Do not present it as a dedicated AI clicks, queries, or conversion report. Its data is also included in the Web search type of the overall Performance report, so do not add the two totals together as separate exposure. Google has a separate Discover report (https://support.google.com/webmasters/answer/16983858?hl=en). Keep those surfaces distinct. Review known reporting anomalies (https://support.google.com/webmasters/answer/6211453?hl=en) when a graph changes unexpectedly. These reports describe Google's supported features. They do not measure ChatGPT, Claude, or every AI answer about your brand. Connect referrals to useful business outcomes OpenAI's publisher FAQ (https://help.openai.com/en/articles/12627856-publishers-and-developers-faq) says ChatGPT adds utm_source=chatgpt.com to referral URLs. Check how your analytics capture that parameter, the referrer, and the destination page. Verify attribution through any redirects you control. In HubSpot, tracking URLs (https://knowledge.hubspot.com/settings/how-do-i-create-a-tracking-url) are intended for links you distribute with campaign parameters. They do not make it possible to attach your own tracking to every link an independent answer system creates. Use the reports and CRM fields your setup actually supports. Review how an identifiable AI arrival is stored, how later form submissions are associated, and what happens across repeat visits. Tracking, consent, and session limitations mean some journeys will remain unattributed. Do not reclassify all direct traffic as AI traffic. Define a meaningful conversion before reporting improvement. A submitted enquiry, a qualified enquiry, and a won deal answer different business questions. Where volumes allow, compare lead quality and progression, not only visits. A self-reported “how did you hear about us?” answer can supplement the evidence, but keep it distinct from recorded referral attribution. Turn observations into a clear next action Match the evidence to the work: - Failed verified requests: investigate crawler access (https://superschema.ai/hubspot-aeo/ai-bot-tracking). - Incorrect service descriptions: clarify the public service page and supporting business facts. - Links to weak evaluation pages: improve the information those pages provide. - Visits without qualified enquiries: inspect offer fit, page clarity, and the enquiry journey. - Conflicting organization details: review identity consistency and Organization schema (https://superschema.ai/hubspot-aeo/organization-schema). Keep a dated change log. If content, internal links, schema, security settings, and campaigns all change together, a later improvement cannot be confidently assigned to one change. Report the sequence as an observation unless your evaluation design supports a stronger conclusion. A useful monthly summary contains the coverage, sample results, accuracy issues, identifiable referrals, qualified conversions, and one next action. Show uncertainty plainly. Better measurement should make decisions clearer, even when the evidence does not yet show improvement. Turn observations into useful work Use the findings to choose the next improvement to your HubSpot website. See how SuperSchema helps HubSpot websites (https://superschema.ai/hubspot-aeo) Sources Reviewed September 29, 2026. - Google: Generative AI performance report, Search (https://support.google.com/webmasters/answer/16984139?hl=en). - Google: Generative AI performance report, Discover (https://support.google.com/webmasters/answer/16983858?hl=en). - Google: Data anomalies in Search Console (https://support.google.com/webmasters/answer/6211453?hl=en). - OpenAI: Publishers and Developers FAQ (https://help.openai.com/en/articles/12627856-publishers-and-developers-faq). - HubSpot: Create tracking URLs (https://knowledge.hubspot.com/settings/how-do-i-create-a-tracking-url).