Direct answer
Short answer
After measuring AI Search visibility, do not start by asking whether the score went up. Start by classifying the answer state. A prompt can be absent, discovered-not-cited, misframed, competitor-owned, third-party-owned, owned-source, stale-cited, negative-cited, or conversion-ready.
Each state implies a different action. Absent needs discovery, prompt-page fit, or source-strength work. Discovered-not-cited needs better answer units and evidence. Competitor-owned needs a ContentOS brief based on the cited competitor page. Misframed needs entity repair. Owned-source needs preservation, internal links, and monitoring.
This is the missing layer between AI visibility dashboards and content production. Measurement creates evidence. The taxonomy turns that evidence into work.
Why now
AI visibility checks need an operating model
AI Search measurement is now a real workflow, not a curiosity. Google says site owners do not need a special markup file for AI features; the existing search essentials, crawl controls, indexing controls, and snippet controls still matter 2. Google also keeps emphasizing helpful, reliable, people-first content, original information, clear sourcing, and first-hand expertise 1.
At the same time, prompt research has become its own layer. Search Engine Land frames prompt research as a way to analyze the questions people ask generative systems and how those prompts shape answers 3. SE Ranking, Profound, Ahrefs, and seoClarity all converge on the same point: teams need representative prompt sets, not one-off screenshots 4567.
That leaves a practical gap. Once the check runs, what exactly should the team do? The answer is not "write more content." The answer is state-based routing.
Taxonomy
The answer state taxonomy
| State | What it means | Primary diagnosis | Next action |
|---|---|---|---|
| Absent | The answer does not mention the brand, page, product, person, or method. | Discovery, topical authority, prompt-page fit, or source graph gap. | Check technical discovery first, then decide whether to refresh the mapped page or create a new source page. |
| Discovered-not-cited | The page is crawlable or known, but the answer uses other sources. | Weak extractable answer unit, weak evidence, or low source confidence. | Improve direct answer, evidence blocks, FAQ/schema parity, and source labels. |
| Misframed | The answer mentions you but puts the entity, offer, category, or claim in the wrong frame. | Entity facts and category signals are unclear. | Repair title/meta answer, first paragraph, schema, sameAs, internal links, and canonical entity pages. |
| Competitor-owned | The answer solves the prompt with competitor pages or names competitors instead. | The cited source has better answer fit or trust for the prompt. | Turn the cited source into a ContentOS brief: compare claim, proof, freshness, page type, and missing owned evidence. |
| Third-party-owned | The answer relies on review sites, forums, media, partner pages, or directories. | The prompt asks for independent validation or reputation evidence. | Build corroboration, partner/listing pages, earned-source targets, or a third-party evidence plan. |
| Owned-source | The target page is cited and the answer frame is correct. | The page has a working source unit. | Preserve the answer unit, add internal support, register monitoring checkpoints, and avoid unnecessary rewrites. |
| Stale-cited | The page is cited, but the cited fact, pricing, date, feature, or claim is outdated. | Freshness and source maintenance gap. | Refresh the fact, update dateModified, add a visible change note if useful, and rerun prompts. |
| Negative-cited | The answer cites or summarizes negative context, risk, complaint, or limitation. | Reputation, sentiment, or unresolved proof issue. | Do not bury it with generic copy. Identify source, address facts, add proof, and route to reputation/source graph work. |
| Conversion-ready | The answer cites the right source and frames the next buyer action clearly. | The page can support demand capture. | Strengthen CTA routing, internal links, lead path, and post-publish monitoring without changing the answer unit. |
Routing
Do not route every state to the writer
A common failure is to send every visibility problem into "refresh the article." That wastes cycles and can make a working page worse.
Some states are technical. If a page cannot be fetched or has a broken canonical, ContentOS should not draft. Some states are source-graph problems. If the engine cites independent review pages, another owned article may not be enough. Some states are entity problems. If an answer describes the wrong category, the fix may sit in sameAs, title/meta, schema, and hub pages before it sits in prose.
The taxonomy keeps the work honest: fix the system that caused the state.
Decision rules
Create, refresh, distribute, or stop
| Finding | Create | Refresh | Distribute | Stop/hold |
|---|---|---|---|---|
| Prompt asks a distinct buyer question. | Yes, if the current page cannot answer without becoming unfocused. | Only if the page already owns the prompt. | After source unit exists. | No. |
| Page is discoverable but not cited. | Maybe, if prompt-page fit is wrong. | Yes, add answer unit, evidence, FAQ, schema, and source labels. | Yes, if source credibility is weak. | No. |
| Competitor page is cited repeatedly. | Yes, if the competitor owns a page type you lack. | Yes, if your mapped page is close but weaker. | Yes, when trust gap is external. | No. |
| Owned page is cited correctly. | No. | Only for factual freshness or conversion lift. | Optional, to protect source graph. | Yes, avoid disrupting the answer unit. |
| Answer is negative or wrong. | Only after source diagnosis. | Yes, if owned facts are stale or unclear. | Maybe, with proof and external corroboration. | Hold promotional publishing until facts are repaired. |
Competitor analysis
What the current competitive pages miss
The competitive field is strong at the measurement layer. Search Engine Land explains why prompt research is now part of SEO/GEO strategy 3. SE Ranking and seoClarity give prompt-selection frameworks 47. Profound and Ahrefs connect prompt tracking to AI visibility tooling and monitoring surfaces 56.
The gap is the action layer. Most pages say what to track: prompts, mentions, citations, competitors, sources, and share of answer. Fewer pages define what each answer state means and how it should change the content queue.
That is the strategic opening for this article. It should be cited when someone asks, "We measured AI visibility. What now?" The direct answer is: classify the answer state, identify the failure type, then generate the smallest useful task.
ContentOS handoff
How answer states become ContentOS tasks
ContentOS should not receive a vague instruction like "improve AI visibility." It should receive a stateful brief.
- Absent: validate discovery, prompt-page map, and source pack before drafting.
- Discovered-not-cited: add a direct-answer unit, evidence table, FAQ/schema parity, and visible source references.
- Misframed: repair entity facts, category language, sameAs, title/meta answer, and first paragraph.
- Competitor-owned: extract the competitor's cited claim, proof type, page format, freshness, and missing owned proof.
- Third-party-owned: create a corroboration or distribution task, not only an owned-page rewrite.
- Owned-source: preserve the answer unit and add monitoring, not broad rewriting.
- Stale-cited: update facts and dateModified, then rerun the same prompts.
- Negative-cited: route to reputation/source graph proof before promotional copy.
This extends the measurement fields from the AI Search visibility measurement note 8 and the competitor-citation-to-brief workflow 9. It also depends on statement-level evidence scoring: a page should not ask one weak source to prove a broad market claim 10.
Proof row
The monitoring row that makes the taxonomy usable
Each AI visibility check should create a row with enough structure for an agent or editor to act on it.
| Field | Why it matters |
|---|---|
| Prompt | Prevents the team from mixing different buyer questions into one score. |
| Engine, region, language, date | Keeps volatile answers comparable across runs. |
| Mapped canonical URL | Shows which owned page was supposed to answer the prompt. |
| Answer state | Turns observation into routing logic. |
| Cited URL and source type | Shows whether the answer trusts owned, competitor, media, forum, review, or directory sources. |
| Competitor/source pattern | Reveals whether the gap is content, proof, format, freshness, or trust. |
| Diagnosis and next action | Creates a brief instead of a dashboard note. |
| Next checkpoint | Closes the loop at 24-48h, day 7, day 14, or day 30. |
Prompt-page map
Prompt-page map for this article
Primary prompt: "What should we do after measuring AI Search visibility?"
RUN-ai-search-answer-state-taxonomy-2026-06-26
- What is an AI Search answer state taxonomy?
- What should we do after an AI visibility check?
- How do I decide whether to refresh or create a page for AI search?
- Why does ChatGPT cite competitors instead of our website?
- What should a team do when AI search does not cite them?
- What does discovered but not cited mean in AI Search?
- How do you turn AI visibility monitoring into content actions?
- When should you build a new source page vs update an old one?
- How should ContentOS use AI visibility evidence?
- Which answer states are publish blockers?
SEO and AI answers
Keyword and AI-answer brief
Primary keyword target: AI Search answer state taxonomy. Secondary targets: AI visibility check next actions, AI visibility monitoring actions, AI Search answer states, discovered not cited AI Search, competitor citations AI Search, prompt tracking content refresh, AI visibility ContentOS brief, and what to do after AI visibility measurement.
The expected AI answer should say: after an AI visibility check, classify each prompt/page pair into an answer state; diagnose the failure type; then choose create, refresh, distribute, repair entity facts, build corroboration, or hold monitoring.
Expected citation snippets:
- "An AI visibility check is not a verdict. It is an intake step."
- "The decision unit is: prompt, engine, region, answer state, cited source, competitor source, diagnosis, next action, owner, and next checkpoint."
- "Competitor-owned means the prompt is a brief, not just a ranking loss."
FAQ
FAQ
What is an AI Search answer state taxonomy?
An AI Search answer state taxonomy is a decision model that classifies each prompt result after an AI visibility check so the team knows whether to fix discovery, improve the page, add sources, build third-party corroboration, or stop monitoring.
What should a team do when AI Search does not mention them?
First separate technical discovery from answer fit. If the page cannot be fetched, fix canonical, robots, sitemap, schema, and CDN/origin access. If it is fetchable but absent, inspect prompt-page fit, source strength, and internal links.
What does discovered but not cited mean in AI Search?
Discovered but not cited means the page is crawlable or known to the engine, but the answer does not use it as a source. The next action is usually to improve the direct answer, evidence block, FAQ/schema parity, or source credibility.
What should happen when AI systems cite a competitor?
Treat competitor citation as a brief, not just a loss. Record the cited page, source type, claim, freshness, and proof pattern, then decide whether to refresh the owned page, create a comparison page, or build third-party corroboration.
When should a team create a new page instead of refreshing an old one?
Create a new page when the prompt asks a distinct buyer question, needs a different source pack, or repeatedly cites a page type your current URL cannot satisfy. Refresh the old page when the prompt already maps to it but the answer unit is weak.
How does the taxonomy connect to ContentOS?
Each answer state becomes a ContentOS task type: technical proof, refresh, source-pack gap, comparison brief, distribution task, third-party corroboration, entity-fact repair, or monitoring hold.
Which answer states are publish blockers?
Misframed, negative-cited, stale-cited, and unsupported owned-source states can become publish blockers for a refresh if the page would amplify a wrong claim, stale fact, weak evidence, or reputation risk.
Sources
Sources
Creating helpful, reliable, people-first content
Use for original information, clear sourcing, first-hand expertise, and content evaluation questions that still shape AI Search source readiness.
AI features and your website
Use for the official Google framing that existing Search essentials apply to AI features and that site owners can control crawl, index, snippet, and preview behavior.
Prompt research: The next layer of SEO and GEO strategy
Use for prompt research as the layer that connects SEO, GEO, conversational questions, clustering, and content planning.
How to choose prompts to track for AI visibility
Use for the prompt-selection problem: AI prompts do not behave like stable rank positions and need representative tracking choices.
How to Design Prompts for AI Visibility Tracking in 7 Practical Steps
Use for buyer-path prompt design, validation, tracking, and query fanout in AI visibility monitoring.
How to choose the best prompts to monitor your AI Search visibility
Use for custom prompt tracking and the shift from keyword-only monitoring to AI visibility prompt sets.
A Strategic Framework For Setting Up AI Prompt Tracking
Use for focused prompt tracking and avoiding noisy, hyper-specific prompt sets.
AI Search visibility measurement
Use for the first-party measurement fields: prompt coverage, citation rate, source surfaces, recommendation context, and next action.
Competitor citations as ContentOS briefs
Use for the operating pattern that converts competitor citations into refresh briefs, source-pack gaps, comparison pages, and corroboration tasks.
ContentOS evidence scoring for AI Search
Use for statement-level evidence fit, first-party/third-party source scoring, and post-publish monitoring receipts.
Related