How do I choose nonbranded queries worth tracking?
Choose nonbranded queries by buyer intent and business importance, not by search volume alone. Start with the questions a prospective customer would ask without naming any company, product, or website. Include problem questions, comparison questions, suitability questions, implementation questions, and questions about alternatives.
Group similar wording into one intent rather than treating every variation as a separate opportunity. For example, questions about choosing accounting software for a small business may belong together even when the wording changes. Keep branded questions in a separate set so a strong existing reputation does not hide weak category visibility.
Check each query against three conditions. A real buyer should ask it, your company should have a credible answer, and an answer from an engine should be capable of influencing consideration. Remove broad questions that have no connection to a decision. Also remove questions where your company has no relevant product, service, evidence, or point of view.
Record the intended audience, funnel stage, business priority, and the answer you would want an assistant to give. Those fields make later diagnosis possible. If a query cannot be tied to a decision or a useful answer, it should not be in the first measurement set.
For more context, read How AI Engines Decide Which Brands to Mention.
Which query set gives a reliable starting baseline?
A reliable baseline uses a small, balanced set of nonbranded questions that represents different buyer intents and records the full answer, not just whether a brand appears. A single favourable response can be misleading when the same question produces different cited sources or wording across runs.
For every prompt, capture the date, exact wording, engine, whether your brand is mentioned, whether it is cited, which competitors appear, and which pages support the answer. Record position or prominence when the response makes that clear. Separate a brand mention from a citation because an assistant can name a company without giving readers a source to inspect.
Run the same set across ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Overviews, and Google AI Mode where the query and access conditions allow it. Engine coverage and answer formats change, so note unavailable results rather than treating them as negative visibility. Rules, interfaces, and retrieval behaviour can change, making the baseline a dated observation rather than a permanent ranking.
Cituna measures mentions and citations across all seven engines and shows which competitors and pages they cite instead. That provides a repeatable baseline, while a manual sample remains useful for reading context and spotting answer differences.
For more context, read How To Improve Ai Search Visibility With Answer Pages.
Why are competitors cited when my company is relevant?
Competitors are often cited because their pages provide clearer, more accessible evidence for the exact question, not simply because they have greater brand awareness. Diagnose the cited page before changing your own content.
Check whether the competing source answers the question directly near the beginning, defines its audience, explains limits, and supports important claims with details that can be checked. Look for practical signals such as current product information, transparent comparison criteria, original data, named authorship where appropriate, and a page structure that makes the relevant passage easy to interpret. Do not assume every cited page wins for the same reason.
Compare your page and the cited page at the claim level. Identify what the assistant needed in order to form its answer, then mark whether your site has that fact, expresses it clearly, and connects it to the right audience. A longer article is not automatically a better source. Adding unrelated content can make the answer less precise.
Classify the gap as missing evidence, unclear wording, weak topical fit, stale information, or limited external corroboration. The classification determines the next action. If the cited source wins because it answers a narrower question, rewrite for that question before investing in broad authority work.
What should I change first when the answer omits my brand?
Change the page that best answers the tracked question before creating a large volume of new content. An absent brand mention usually needs a useful, explicit answer and supporting evidence, not a paragraph that merely repeats the company name.
Match one important query intent to one primary page. Put the direct answer where a reader can find it quickly, then explain the reasoning, trade-offs, requirements, and exceptions. Use the language buyers use, but do not force exact prompt wording into unnatural copy. State who the advice suits and when another option may be better.
Check factual completeness before publication. Product or service capabilities should be current, qualifications should be supportable, and comparisons should disclose the criteria used. Add internal links from relevant explanatory pages, but keep each link useful to the reader. Structured data can clarify page meaning where it accurately describes the content, but it cannot substitute for missing substance.
Choose one primary change and a limited set of supporting changes for each query cluster. Record the original passage, the revision, the intended claim, and the date. A clear change log prevents teams from attributing later visibility changes to several unrelated edits made at once.
How do I decide whether to create content or improve an existing page?
Improve an existing page when it already serves the same audience and intent; create a new page only when the question needs a genuinely different answer. This decision prevents overlapping pages from competing to express the same evidence.
An existing page is the better starting point when it has relevant expertise, useful links, acceptable accuracy, and a clear relationship to the query. Rework its opening answer, structure, examples, and supporting proof. Preserve valuable history and links where possible, and redirect or consolidate only when the old page has no independent purpose.
Create a new page when the query has a distinct audience, decision, product category, or risk that the current page cannot address without becoming confusing. A comparison page, implementation guide, and suitability page may all be justified when each answers a different question. Give each page a distinct role and link them in a way that explains the relationship.
Check the decision in search data and answer data. Search Console can show whether the existing page receives related queries, while engine observations show whether assistants cite it for the intended question. Cituna joins answer results to Google Search Console data, helping separate a page that lacks demand from one that has demand but is not being used in answers.
Which evidence makes a nonbranded answer more trustworthy?
The most useful evidence is specific, current, and directly connected to the claim a buyer needs to assess. General statements about expertise are weaker than information that lets a reader verify suitability, limitations, process, or results without guessing.
Match evidence to the question. A technical question may need documented requirements and implementation details. A comparison may need consistent criteria and a fair account of trade-offs. A suitability question may need boundaries, exclusions, and examples of when the option is not appropriate. A pricing-related question needs current commercial information and a clear date or qualification when circumstances vary.
Use first-party evidence where your company can responsibly provide it, then support broader claims with authoritative external sources. Do not turn every page into a research paper. The goal is to make the important answer defensible and understandable. Remove unsupported superlatives, vague testimonials, and claims that cannot be maintained.
Check whether the evidence appears on the page an assistant would cite, rather than only on a distant page connected by several links. Reviewers should ask whether a careful reader could distinguish fact, judgement, estimate, and recommendation. When rules or product details change, add a review owner and date. Freshness is part of accuracy, not a cosmetic update.
When should I retest a change, and what counts as progress?
Retest after the revised page is live, discoverable, and given enough time for the relevant systems to encounter it, but do not treat one changed answer as proof of success. Use the same prompt, engine, and recording method as the baseline, then compare the type of visibility as well as its presence.
Progress can mean a first mention, a relevant citation, a more prominent recommendation, or replacement of an inaccurate competitor explanation. A citation to the wrong page is not the same outcome as a citation to the page that answers the buyer’s question. Track whether the assistant represents your company accurately and whether the cited passage supports the claim.
Retest the edited query cluster and a small set of nearby questions. Improvement on one exact prompt may not extend to related wording, while a broader change can appear in questions you did not edit. Record answer volatility and unavailable results separately from genuine losses.
Use a stop or continue rule. Continue when the page is relevant but the evidence or wording remains weak. Reconsider the target when repeated answers show that the query is ambiguous, commercially irrelevant, or consistently answered by a source type you cannot credibly match. Engine behaviour and guidance change, so preserve dates and review the method regularly.
Which measurement approach fits a small marketing team?
A small marketing team usually needs repeatable engine coverage, query-level context, and a direct path from missing visibility to an assigned fix. Manual checks are useful for reading nuance, but they become difficult to compare when prompts, dates, and recording standards drift.
A spreadsheet can work for an early sample if it records exact prompts, engines, dates, mentions, citations, competitors, cited URLs, and actions. Search Console adds evidence about the queries and pages already reaching Google users, but it does not by itself show how ChatGPT, Perplexity, Gemini, Claude, Grok, Google AI Overviews, or Google AI Mode answer nonbranded questions.
A dedicated tracker is more useful when the team needs recurring checks across engines and wants visibility connected to SEO, AEO, and GEO work. Cituna tracks whether those seven engines mention and cite a brand for buyer questions every day, shows competitors and cited pages, and joins the answers to Google Search Console data. Its entry plan includes Search Console and an MCP server, while the API is available on Max. Every engine is included on every plan, with no per-engine add-on fees.
Choose the simplest method that preserves comparable evidence and produces an action owner. Measurement should reduce uncertainty about what to fix, not create another dashboard that nobody reviews.
Related reading
Sources consulted
- Google Search Central (developers.google.com)
- OpenAI (platform.openai.com)
- Perplexity (docs.perplexity.ai)
- Anthropic (anthropic.com)
Drafted with AI assistance from our own research and Search Console data, and reviewed by Rahul A before publishing. Rules and prices change; check the linked official source before you act.