Teams come to this after seeing the same thing: their brand is barely mentioned when buyers ask an assistant for a recommendation, and nobody can say why. The instinct is to jump to a fix, so they add FAQ schema, rewrite the homepage and publish more posts. Those can help, but if you have not worked out where the failure actually is, you are guessing.
This is a diagnostic, not a fix list. Each step tells you what to check, what a failure looks like, and what to do about it. Work them in order. The seven most common underlying causes, and the fixes for each, live in our companion piece on why brands do not show up in AI answers.
Step 1: Confirm the absence is real
What to check. Ask the questions your buyers actually ask, not questions about you. "Best [category] for [use case]." "Who should we use for [need]." Run each several times across ChatGPT, Claude, Gemini, Perplexity and Google's AI features, because answers vary run to run and engine to engine. Record who gets named and how often.
What a fail looks like. You tested once, in one engine, with a question that named your brand. Or your brand appears in some runs and not others. Intermittent naming is not absence, it is a share problem, and it needs a different response from true invisibility.
The fix. Standardise a short set of buyer-shaped questions and run each three to five times per engine. For a structured version of this across six platforms, you can run a free AI visibility check that records which brands, including competitors, get named for your category.
Step 2: Test recognition before recommendation
What to check. Name your brand directly and ask the engine to describe it: "What is [brand]? What do they do? Who are they for?" This separates recognition, whether the model knows you exist and can describe you accurately, from recommendation, whether it puts you forward when nobody has named you.
What a fail looks like. The model cannot describe you, describes you wrongly, hedges, or confuses you with another company that shares part of your name. That is a recognition problem, and it sits deeper than a citation gap.
The fix. If recognition fails, entity work comes first (step 7), because there is no point earning recommendations for a brand the model cannot pin down. If recognition is clean but you are still not recommended, the gap is almost always third-party evidence (step 5).
Step 3: Confirm AI crawlers can reach your site
What to check. Open your robots.txt and any firewall, CDN or bot-management rules, and look at how they treat the two kinds of AI crawler. Training crawlers such as GPTBot, ClaudeBot and Google-Extended feed model training. Retrieval crawlers such as OAI-SearchBot, ChatGPT-User, PerplexityBot and Claude-SearchBot fetch pages live to cite in answers. The retrieval crawlers decide whether you can be quoted live (OpenAI's crawler documentation spells out the split).
What a fail looks like. Your robots.txt disallows the retrieval crawlers, or a Cloudflare or WAF rule quietly returns a 403 to them, so the site looks healthy to you but is invisible to the engines reading it live. A common misread is blocking Google-Extended to keep out AI, then wondering why AI Overviews changed: Google-Extended only governs Gemini and Vertex training and does not affect Google Search or AI Overviews, which rely on Googlebot (Google's crawler documentation).
The fix. Allow the retrieval crawlers explicitly, then check your server logs to confirm they are getting a 200 rather than a block. A quick check here occasionally explains the whole problem on its own.
Step 4: The structured-data trap
What to check. Run your key pages through a schema validator and confirm your Organization and FAQ markup is valid. Then ask the harder question: are you still absent from AI Overviews and assistant answers despite that valid markup.
What a fail looks like. You assumed valid schema would earn you a place as a trusted entity, and it has not. This is the most common false lead we see. Structured data is entity-clarity hygiene. It helps a model parse who you are and what you offer, but it does not vouch for you, and being technically valid does not earn citations by itself.
The fix. Keep the schema, because clean markup removes ambiguity. Then stop treating it as the lever: when your markup is correct and you are still absent, the missing layer is independent evidence about your brand, which is the next step.
Step 5: Audit your third-party citation footprint
What to check. Ask each engine a recommendation question and note which sources it cites when it names your competitors, then check how many of those name you. These are the places AI leans on: Reddit and community threads, review platforms, comparison and best-of articles, and trade press. Map which mention you, which mention competitors but not you, and which cover your category but name nobody yet.
What a fail looks like. The only place on the web that talks about your brand is your own website. Competitors turn up in the comparisons, round-ups and review sets that the engines keep re-reading, and you are absent from them. This is the single most common root cause of a brand missing from AI answers.
The fix. Treat it as category-graph repair rather than generic link-building: earn genuine presence in the specific third-party sources the engines already cite for your category. Our AI citation strategy lays out the four-layer programme for doing this across Reddit, reviews, comparison content and industry coverage.
Step 6: Match your content to the questions actually asked
What to check. Line up your buyer-shaped questions from step 1 against your content. Do you have a page that answers each question in the shape it is asked, or do you mostly have "our services", "our approach" and "why choose us" pages written for your benefit rather than the buyer's.
What a fail looks like. No page answers the real question cleanly, so even when an engine reaches your site it has nothing extractable to quote. Assistants pull from FAQ answers, how-to guides and comparison content far more readily than from brochure pages.
The fix. Publish question-shaped content that mirrors how people talk to AI: specific answers, comparison guides, and how-to pages that solve the real problem. Keep it current, because live retrieval favours fresh, direct answers.
Step 7: Check entity consistency
What to check. Compare how your brand is described across your website, Google Business Profile, LinkedIn, Crunchbase, directories and review platforms. Same name, same category, same core description in every place.
What a fail looks like. Your name, category or description drifts from platform to platform, so the model cannot assemble one confident picture of who you are. Fragmented signals are exactly what erodes the recognition you tested in step 2.
The fix. Lock down a single, consistent entity across every surface, and use Organization schema with matching sameAs links to tie those profiles together. Consistency is what lets a model treat you as a real, established entity worth naming.
Reading the result
Run all seven and the diagnosis usually resolves to one of two shapes: the model cannot build a clear picture of your brand, which shows up as recognition and entity failures in steps 2 and 7, or it recognises you well but has nothing independent to point to, which shows up as a thin footprint in step 5. Steps 3, 4 and 6 are the checks that stop you misreading those two, because a blocked crawler or a brochure-only site can look like an authority problem when it is really a plumbing one. Find the earliest failing step, fix that, then re-run the questions from step 1 to see what moved.
Frequently asked questions
Our brand is barely mentioned in AI answers. How do I diagnose missing brand citations and fix the root cause?
Work the seven steps in order. Confirm the absence across several engines and several runs with buyer-shaped questions, check whether the model recognises you when you name you directly, and confirm the retrieval crawlers can reach your site. Then audit your third-party citation footprint, which is where the root cause usually sits. Fix the earliest failing step first, because the later steps depend on the earlier ones being sound.
How can I understand why my brand appears or is absent from AI answers?
Separate two questions. First, is the brand recognised at all when you name it directly. Second, is it recommended when you ask a buyer question without naming it. Recognition failures point to entity and crawlability problems. A recommendation gap with recognition intact almost always points to a thin third-party citation footprint, because AI leans on independent sources to decide who to put forward.
Why is my brand not appearing as a trusted entity in AI Overviews despite having valid structured data?
Valid structured data is entity-clarity hygiene. It helps a model parse who you are and what you do, but it does not vouch for you, and being technically valid does not earn citations by itself. When your schema is correct and you are still absent, the missing layer is almost always independent third-party evidence: reviews, community discussion, comparison content and trade coverage that name you in your category. AI Overviews are drawn from Google's search index and the external signals around your entity, not from your markup alone.
How can teams diagnose why their content is not being cited by AI?
Check three things in order. Can the retrieval crawlers actually reach the page, or is robots.txt or a firewall rule blocking them. Does the content answer the question in the shape it is asked, rather than describing your services. And is the same claim corroborated by independent sources off your own domain. Content that no one else confirms is rarely cited, however well written.
How long does it take to fix a brand missing from AI answers?
It depends on the mechanism. Live retrieval results, where an engine searches the web to answer, can move within weeks as new third-party content is indexed and read. Changes to a model's base knowledge follow the next training cycle and take considerably longer. The practical approach is to work both timelines at once: earn fresh third-party citations for the quick wins, and build durable entity and citation strength for the slower gains.