Published by Roast & Rise
AI Search Visibility: A Practical Sprint
Audit visibility, publish useful source pages, set crawler policy, and measure what changes.
Audit how your company appears in AI search, publish verifiable source pages, set crawler policy, and measure changes in Search Console.
Course thesis
AI search visibility starts with useful, crawlable, indexable pages that answer real questions and support their claims. For Google's AI features, core Search requirements still apply; special AI files and markup are not ranking shortcuts.
What you leave with
You will leave with a query baseline, a claim-to-source ledger, one publishable page brief, a documented crawler policy, and a measurement decision.
For
Founders, operators, marketing leads, product owners, and web teams responsible for public company knowledge and organic discovery.
Workflow
Benchmark priority queries, map claims to primary sources, improve one page, verify Search controls, and review the evidence on a fixed cadence.
Change
Teams replace untracked AI-search experiments with an owned publishing and measurement loop tied to specific queries, pages, and evidence.
What you can do
Use these as checks while you move through the plan.
Build a repeatable AI search visibility baseline for priority queries and pages.
Connect each important public claim to a strong, primary source.
Brief pages that answer the query directly and expose supporting evidence.
Separate Google Search controls from policies for other AI uses and crawlers.
Use Search Console and dated answer checks to decide what to keep, change, or stop.
Chapters
01
Audit AI Search Visibility
Establish which non-brand queries expose your company, which page Google associates with each query, and where the evidence is missing.
AI search visibility is whether a useful page from your company appears, supports, or is cited in an AI-generated search answer for a relevant query. Start with repeated non-brand demand, because brand searches mostly measure existing awareness.
For Google's AI features, a page must meet normal Search technical requirements and be eligible for indexing and snippets. Google says its core Search systems still apply; there is no separate AI ranking shortcut. Source: Google Search Central, Optimizing for generative AI features.
Export the same date range for Search Console queries and pages. The two exports are separate aggregates, so do not claim that a query produced impressions for a specific page unless you obtained query-by-page data from Search Console or its API.
Build the baseline from recurring, genuine queries. Exclude brand terms, obvious test prompts, quoted audit strings, and single-impression noise. Record impressions, clicks, position, the likely intent, the current landing page, and the content gap.
Worked example
Illustrative case: repeated queries ask how to build an AI business case. The site has a broad AI strategy page but no direct ROI decision framework. Mark the intent as partly answered and refresh the closest existing page before creating another URL.
Quality checklist
The date range is identical across exports.
Brand, synthetic, and single-impression noise are excluded and documented.
Variants with the same intent are clustered before prioritization.
Position is weighted by impressions rather than averaged by row.
Existing content is checked before a new URL is proposed.
Common mistakes
Treating every wording variant as a separate content opportunity.
Joining separate query and page exports as if they shared row-level keys.
Prioritizing a single strange impression over repeated demand.
Creating a new page when a relevant published page needs a focused refresh.
Checkpoint
Can another person reproduce the priority list from your exports, exclusions, clustering rules, and content inventory?
Exercise
Build an AI Search Visibility Baseline
Export 12 months of Search Console query and page data. Group repeated non-brand queries by shared intent, then inspect the current site for an exact or partial answer.
For each cluster, record:
- Query cluster
- Included query variants
- Impressions and clicks
- Impression-weighted position
- Current page, if known
- Intent answered fully, partly, or not at all
- Evidence needed
- Candidate page action
- Owner
Output to complete
AI Search Visibility Baseline
Copyable template
Query cluster:
Included query variants:
Impressions:
Clicks:
Impression-weighted position:
Current page:
Intent coverage: full/partial/absent
Evidence needed:
Candidate action:
Owner:
Use this at work tomorrow
Export the last 12 months of Search Console data and classify the five largest clean non-brand intent clusters before proposing new content.
Core idea
A useful AI search plan starts with clean query demand and an honest content inventory.
02
Map Claims to Primary Sources
Connect each claim a priority page makes to current, public evidence and mark the gaps that require research or removal.
A claim is usable only when a reader can see what it means, where the evidence comes from, and what limits apply. Unsupported superlatives, anonymous statistics, and copied summaries weaken the page.
Prefer primary and official sources for facts about standards, policies, laws, and platform behavior. Use first-party company evidence only when its method and scope can be explained without exposing confidential information.
Google recommends unique, non-commodity content that reflects real expertise and experience. Repackaging what already exists adds little value. Source: Google Search Central, Optimizing for generative AI features.
The Claim and Source Ledger records the exact claim, source label and URL, evidence type, scope, limitation, page using the claim, accountable owner, and action: keep, qualify, replace, or remove.
Worked example
Illustrative case: a page says an AI workflow is more accurate but provides no task, baseline, or evaluation method. The ledger marks the claim for removal and replaces it with a description of the tested workflow and its documented limits.
Quality checklist
Every claim is specific enough to check.
Primary or official sources support platform, legal, and standards claims.
Source labels tell readers who published the evidence.
Scope and limitations appear beside material claims.
Unsupported claims have a named action and owner.
Common mistakes
Citing a search result, aggregator, or secondary summary when a primary source exists.
Keeping a precise statistic after its method or scope can no longer be verified.
Linking a source that discusses the topic but does not support the exact claim.
Adding a review date that no accountable person actually performed.
Checkpoint
Can a skeptical reader trace every material claim on the priority page to relevant evidence and see its limits?
Exercise
Build the Claim and Source Ledger
Choose one priority page and inspect every factual or comparative claim. Link each claim to the strongest available primary source or mark it for qualification or removal.
Use these columns:
- Exact claim
- Page and section
- Source label
- Source URL
- Evidence type
- Scope and limitation
- Accountable owner
- Action: keep, qualify, replace, or remove
Output to complete
Claim and Source Ledger
Copyable template
Exact claim:
Page and section:
Source label:
Source URL:
Evidence type:
Scope and limitation:
Accountable owner:
Action: keep/qualify/replace/remove
Use this at work tomorrow
Audit the five strongest claims on one priority page and remove or qualify any claim that lacks defensible support.
Core idea
If a claim cannot survive a source check, qualify it, replace it, or remove it.
03
Publish Useful Source Pages
Turn one clean query intent into a page that gives a direct answer, shows evidence, covers failure modes, and helps the reader act.
A useful source page answers the searcher's main question in its opening section, then explains the method, inputs, decisions, failure modes, and limits needed to apply the answer.
Write for the shared intent behind a query cluster. Google advises against creating separate pages for every possible query variation and says no AI-specific rewriting, schema, or chunking is required. Source: Google Search Central, Optimizing for generative AI features.
Give the page a descriptive title, visible text, relevant internal links, and structured data only when it accurately matches the visible content. Keep key evidence in accessible HTML rather than relying on a slide deck or gated asset.
The Useful Source Page Brief names the direct answer, audience decision, primary evidence, practical steps, failure modes, exclusions, internal links, and accountable owner before drafting begins.
Worked example
Illustrative structure for an AI-ready data page: define readiness for one workflow, map data sources, test fitness and access, list blocking failure modes, assign fixes, and set monitoring ownership. Avoid a generic data maturity overview.
Quality checklist
The opening section answers the primary intent directly.
The outline reflects one intent cluster rather than one page per wording variant.
Primary evidence is linked and material limits are visible.
The page contains a usable checklist, template, or decision aid.
Relevant published pages have an internal-link plan.
Common mistakes
Opening with a broad market trend instead of the answer.
Writing a commodity summary that adds no first-hand method or judgment.
Creating several thin pages for close query variants.
Adding schema that describes content the reader cannot see.
Publishing without an accountable subject-matter owner.
Checkpoint
Could a reader use the brief to make the target decision without asking what the page is really about?
Exercise
Brief One Useful Source Page
Choose the highest-priority intent that current content answers only partly. Draft the page brief before writing.
Include:
- Primary query intent
- Audience and decision
- Direct answer
- Supporting questions
- Primary evidence and links
- Practical steps or checklist
- Failure modes
- Limits and exclusions
- Existing pages that should link here
- Pages this page should link to
- Accountable owner
Output to complete
Useful Source Page Brief
Copyable template
Primary query intent:
Audience and decision:
Direct answer:
Supporting questions:
Primary evidence and links:
Practical steps or checklist:
Failure modes:
Limits and exclusions:
Internal links in:
Internal links out:
Publish location:
Accountable owner:
Use this at work tomorrow
Write the opening answer and section outline for one partial content gap, then route it to the person who owns the underlying subject matter.
Core idea
One direct, useful source page can serve an intent cluster better than a stack of thin variants.
04
Set Search and AI Crawler Policy
Verify ordinary Google Search access first, then document separate choices for snippets, model training, grounding, and other crawlers.
Googlebot controls crawling for Google Search, including Google's AI search features. A priority page also needs index and snippet eligibility. Blocking Googlebot can remove the page from the Search systems you are trying to reach.
Google-Extended is a robots.txt product token, not a separate HTTP user agent. It controls use for future Gemini model training and grounding in Gemini Apps and Vertex AI. Google states that it does not affect inclusion or ranking in Google Search. Source: Google Crawling Infrastructure, Google's common crawlers.
Google Search ignores llms.txt and other special AI files. They neither improve nor reduce Google Search visibility. Keep llms.txt only when another system or publishing policy needs it. Source: Google Search Central, Optimizing for generative AI features.
Build the Crawler Policy Map per URL group and purpose. Record Googlebot access, index and snippet controls, the Google-Extended choice, any separately documented crawler rules, the business rationale, owner, and verification method.
Worked example
Illustrative case: a public guide is allowed in robots.txt but carries a noindex header from an old staging rule. The team fixes the header and verifies indexing separately; changing Google-Extended would not solve the Search problem.
Quality checklist
Googlebot access and index eligibility are verified separately.
Canonical and snippet controls match the intended public page.
Google-Extended is not described as a Google Search ranking control.
llms.txt is not presented as a Google Search requirement.
Every separate crawler decision cites current official documentation and has an owner.
Common mistakes
Allowing Googlebot in robots.txt while a noindex directive blocks indexing.
Telling teams to allow Google-Extended to rank in AI Overviews or AI Mode.
Treating llms.txt as a Google ranking tactic.
Copying another crawler's user-agent rule without checking its official documentation.
Changing production controls without a URL-level verification plan.
Checkpoint
Can the web owner explain which control affects Google Search and which choices govern separate AI uses?
Exercise
Create the Crawler Policy Map
Map the priority content paths and inspect robots.txt, page-level robots directives, X-Robots-Tag headers, canonical tags, sitemap inclusion, and internal links. Consult each non-Google system's current official documentation before adding a rule.
For each row, record:
- URL or path
- Content purpose
- Googlebot access
- Index eligibility
- Snippet control
- Canonical URL
- Google-Extended choice
- Other crawler rule and official documentation
- Business rationale
- Verification method
- Owner
Output to complete
Crawler Policy Map
Copyable template
URL or path:
Content purpose:
Googlebot access:
Index eligibility:
Snippet control:
Canonical URL:
Google-Extended choice:
Other crawler rule and official documentation:
Business rationale:
Verification method:
Owner:
Use this at work tomorrow
Verify Googlebot access, index eligibility, canonical URL, and snippet controls for the one page tied to your highest-priority query cluster.
Core idea
Googlebot affects Google Search. Google-Extended and llms.txt do not control Google Search ranking.
05
Measure and Decide What to Change
Combine Search Console evidence with dated answer checks, then make an explicit keep, change, or stop decision for each page action.
AI search visibility is volatile, so one manual result is evidence of that moment, locale, account state, and query wording. Save the date, query, surface, answer, and shown sources when a manual check matters.
Google directs site owners to Search Console's Generative AI performance reporting for Google Search and Discover, alongside established Search performance data. Use first-party Search Console evidence before declaring a page improvement successful. Source: Google Search Central, Optimizing for generative AI features.
Define the review rule before publishing. Keep the change when the page becomes more useful and relevant signals hold. Change it when the intent is still only partly answered or the wrong page is surfaced. Stop the tactic when it adds maintenance without better evidence or reader value.
The AI Search Measurement Scorecard records the baseline, publishing change, Search Console metrics, manual evidence where useful, confounders, decision, owner, and next action. It prevents a ranking claim from outrunning the data.
Worked example
Illustrative review: impressions rise after a refresh, but the comparison period also includes a demand spike. Mark the result inconclusive, keep the useful page, and extend measurement instead of claiming the refresh caused the increase.
Quality checklist
Baseline and comparison windows use the same dimensions and filters.
Search Console evidence is separated from manual answer observations.
Material confounders are written beside the result.
The decision rule was set before interpreting the outcome.
Every row ends with keep, change, or stop plus an owner.
Common mistakes
Comparing different date ranges or filters.
Attributing every movement to the content change.
Treating one personalized answer as a stable ranking.
Ignoring a wrong landing page because aggregate impressions increased.
Continuing a tactic with no named reader benefit or measurable signal.
Checkpoint
Does each page action end in a defensible keep, change, or stop decision with the evidence attached?
Exercise
Run the Measurement Review
Choose a fixed review window appropriate to the site's crawl rate and current demand. Compare like-for-like Search Console periods and annotate material site, seasonality, campaign, or platform changes.
For each action, record:
- Query cluster
- Target URL
- Baseline period
- Change published
- Comparison period
- Impressions, clicks, CTR, and position
- Generative AI report signal, where available
- Dated manual answer evidence, where useful
- Confounders
- Decision: keep, change, or stop
- Owner and next action
Output to complete
AI Search Measurement Scorecard
Copyable template
Query cluster:
Target URL:
Baseline period:
Change published:
Comparison period:
Search Console metrics:
Generative AI report signal:
Dated manual evidence:
Confounders:
Decision: keep/change/stop
Owner and next action:
Use this at work tomorrow
Record the baseline period and decision rule for one page before the next content change ships.
Core idea
Measure page actions against a baseline, record confounders, and make the decision explicit.
30-day path
Week 1: Clean the Search Console export, cluster repeated non-brand intents, and check current content coverage.
Week 2: Audit claims on the priority page and complete the Useful Source Page Brief.
Week 3: Publish the focused refresh and verify Googlebot, index, canonical, snippet, and internal-link controls.
Week 4: Capture the baseline, schedule the comparison window, and record the first keep, change, or stop decision.
Success signals
Share of retained query impressions assigned to a clear intent cluster and content action.
Material claims on the priority page linked to primary evidence or removed.
Priority URL verified as crawlable, indexable, canonical, snippet-eligible, and internally linked.
Search Console comparison recorded with consistent filters and documented confounders.
Each reviewed page action ends with a keep, change, or stop decision and owner.
Reflection prompts
Which query cluster lacks content, and which one is served by an underperforming page?
Which material claim should be qualified or removed before more traffic reaches the page?
What evidence would change your keep, change, or stop decision?
Manager checklist
Approve the exclusions and intent clusters before content production starts.
Require primary sources and visible limitations for material claims.
Keep Google Search controls separate from other AI-use policies.
Define the comparison window and decision rule before publishing.
Assign an owner to every keep, change, or stop action.
In this library
Related RisePlans
Sales Workflow Upgrade
Turn sales research, discovery notes, proposals, follow-up, and handoffs into a repeatable workflow where AI supports preparation and quality without replacing human judgment.
Customer Support Workflow Upgrade
This plan shows founders and support leads how to rebuild their workflow for sharper ticket triage, clearer AI-human collaboration, and seamless handoffs. Get hands-on tools to reduce guesswork, protect your tone, and deliver better support at scale.
AI Sprawl Cleanup Sprint
Bring order to fragmented AI use. Map every tool, prompt, and cleanup hour. Diagnose waste and workslop. Set the rules and rhythms that make team experiments compound instead of costing trust, time, and money.
Want this shaped around your company?
Risey can research your company foundation first, then build a version of this path around your real workflows, customers, and culture.
Start with your company