Quick answer
An AI visibility audit should not end with a scorecard.
The useful output is a short, evidence-backed action queue: which prompts matter, where the brand appears, where it is absent, which sources shape the answer, what competitors are winning, which facts are wrong, who owns the fix, and when the same prompt group will be re-measured.
Use this audit sequence:
- Define the audit scope: platforms, market, language, brand entities, competitors, and prompt groups.
- Capture the baseline: raw answers, mentions, rank, citations, Share of Voice, sentiment, and source URLs.
- Separate branded accuracy from unbranded discovery.
- Read the cited and uncited source trail before assigning work.
- Compare competitor wins by source type, not just by score.
- Route each gap to owned-page, technical, source, PR, profile, product-fact, or positioning work.
- Ship the smallest credible fix and re-run the same prompt group.
For ReachLLM teams, the audit is a loop: observe the answer, diagnose the source path, ship the fix, and prove whether the next answer changed.
What the Ahrefs source gets right
The scheduled source for this run is Ahrefs' AI visibility audit guide. It is useful as a market signal because it treats AI visibility as more than one metric. The guide starts with audit scope, then moves through brand visibility, branded answer accuracy, unbranded topic associations, cited pages, influential mentions, competitor comparison, and action planning.
That order is sensible. A team cannot explain movement until it knows what it measured.
The ReachLLM addition is the execution layer. The audit should not only answer "Are we visible?" It should answer:
| Audit question | Better operating question |
|---|---|
| Are we mentioned? | Which buyer prompt group mentions us, and where are we absent? |
| Are we cited? | Which owned or third-party source made that citation possible? |
| Are competitors visible? | Which source type helped them win: owned page, roundup, directory, review, news, community, or docs? |
| Is sentiment positive? | Which answer text, source, or stale fact created the sentiment? |
| Did the score move? | What shipped before the movement, and can we re-run the same prompt group? |
If the audit does not create a named next action, it is only reporting.
Step 1: define the scope before collecting answers
Start by writing the audit scope in plain language.
| Scope field | Decision to make |
|---|---|
| Platforms | ChatGPT Search, Google AI Overviews, Google AI Mode, Perplexity, Gemini, Claude, Copilot, or another buyer-used surface. |
| Market | Country, language, region, and whether results should be global or local. |
| Entities | Brand name, product names, sub-brands, founders, executives, abbreviations, and common misspellings. |
| Competitors | Direct competitors that belong in Share of Voice and rank comparisons. |
| Prompt groups | Branded fact, category discovery, comparison, capability, trust, pricing, implementation, and local prompts. |
| Time window | Baseline date, refresh cadence, and the re-measurement date after fixes ship. |
Do not run an audit against a vague category like "AI visibility" and call the result precise. If the scope changes, label the new baseline.
Step 2: collect answer evidence, not just metrics
The baseline should preserve the actual answer a buyer would see.
For each priority prompt, capture:
| Evidence | Why it matters |
|---|---|
| Exact prompt | Small wording changes can produce different answer sets. |
| Platform and mode | ChatGPT with web search, Google AI Overview, Perplexity, Gemini, and other surfaces can cite differently. |
| Raw answer | Summaries hide rank order, phrasing, errors, and caveats. |
| Brand mention | Shows whether the brand appears at all. |
| First mention position | Separates leading recommendations from buried mentions. |
| Competitors | Shows which alternatives the buyer sees beside the brand. |
| Citations and source URLs | Reveals the evidence path behind the answer. |
| Sentiment and accuracy | Flags wrong facts, stale claims, weak differentiation, or negative framing. |
| Run date | Needed for before-and-after validation. |
Google's AI feature guidance says normal Search fundamentals still apply for Google AI Overviews and AI Mode, with no special AI-only markup required for inclusion (Google Search Central). So keep technical eligibility in the audit, but do not let it replace answer review.
Step 3: separate branded accuracy from unbranded discovery
Branded prompts and unbranded prompts diagnose different problems.
| Prompt type | What it checks | Example failure |
|---|---|---|
| Branded fact | Whether AI systems understand the company, products, leadership, pricing, or positioning. | The answer uses old pricing or describes the wrong audience. |
| Category discovery | Whether the brand appears when a buyer asks for options before knowing the brand. | Competitors appear but the brand is absent. |
| Comparison | Whether the answer positions the brand fairly against alternatives. | Competitor pages shape the comparison and omit key product facts. |
| Capability | Whether the product is associated with the job it actually performs. | The brand appears for "monitoring" but not "source intelligence" or "execution." |
| Trust | Whether proof, reviews, citations, and third-party context support recommendation. | The answer names the brand but does not recommend it confidently. |
A branded prompt win can hide an acquisition problem. A category prompt loss can hide a perfectly healthy branded source of truth. Report them separately.
Step 4: read citations as a source trail
Citations are not trophies. They are clues.
When a platform exposes sources, classify each cited URL:
| Source type | What to inspect | First useful fix |
|---|---|---|
| Owned product page | Does it answer the prompt directly and cite proof? | Rewrite the page section, add examples, improve internal links, and clarify entity facts. |
| Owned blog post | Is it educational but disconnected from the product? | Add a concise product-fit section and link to the right commercial page. |
| Third-party roundup | Is the brand omitted, miscategorized, or mentioned weakly? | Pursue legitimate inclusion or publish a stronger comparison page if the source is unreachable. |
| Directory or profile | Are categories, descriptions, pricing, screenshots, and links current? | Correct the profile where allowed and align public facts. |
| Review or community page | Does sentiment reflect a real issue, stale data, or unsupported claim? | Fix the underlying product/support issue first, then correct reachable facts. |
| News or analyst source | Does the source validate a competitor's category position? | Build a relevant PR or expert-source angle, not a thin copycat article. |
OpenAI's ChatGPT Search help says search answers may include source links, and it also cautions that web results and citations can be imperfect or incomplete (OpenAI Help Center). Treat citations as evidence to inspect, not as final truth.
Step 5: find competitor gaps by source type
Do not stop at "Competitor X has higher Share of Voice."
Ask which evidence path helped them:
| Competitor pattern | Likely meaning | Better next action |
|---|---|---|
| Competitor is cited from its own page | Their page answers the prompt more directly. | Improve the equivalent owned page before chasing off-site mentions. |
| Competitor is cited from many roundups | The category is shaped by third-party lists. | Identify legitimate sources where the brand belongs and build a proof-led outreach plan. |
| Competitor appears without citation | The model may have stronger category association for them. | Tighten category language across core pages and public profiles. |
| Competitor is recommended for one capability | Their proof is clearer for that job. | Add specific use cases, screenshots, customer context, and claims the source can support. |
| Competitor wins only in one platform | Platform-specific retrieval or index behavior may be involved. | Check that platform's source set, market scope, and access controls. |
This keeps the audit from becoming a generic content calendar. The fix should match the evidence path.
Step 6: check technical eligibility without turning it into a shortcut
Technical checks matter because important pages need to be accessible, indexable, and easy to understand.
Audit:
| Technical check | Question |
|---|---|
| Robots and crawler access | Are key pages unintentionally blocked from search or relevant AI crawlers? |
| Indexability | Are commercial, docs, comparison, and source-of-truth pages free of accidental noindex rules? |
| Canonicals | Does each important page point to the intended canonical URL? |
| Sitemap | Are priority source pages discoverable and current? |
| Visible text | Are key facts present in HTML text, not only images or client-only widgets? |
| Structured data | Does schema match visible page content and clarify real facts? |
| Internal links | Can users and crawlers reach the page from related content? |
OpenAI's crawler documentation says OAI-SearchBot is used for surfacing and linking to websites in SearchGPT and ChatGPT search, and that sites can use robots.txt controls (OpenAI crawler docs). That is eligibility hygiene. It does not guarantee a citation or recommendation.
Google's robots meta documentation also makes clear that snippet and index controls affect how Search can show content (Google Search Central). The audit should inspect those controls specifically rather than searching the whole page for scary words.
Step 7: route every gap to one owner and one fix
An AI visibility audit becomes useful when each gap has one owner.
| Gap | Owner | First fix |
|---|---|---|
| Brand absent from category prompts | Product marketing or content lead | Improve category and comparison pages with answer-first sections and proof. |
| Brand mentioned but not cited | Web/content owner | Strengthen the crawlable owned source page and internal links. |
| Wrong facts or stale pricing | Product marketing, product, or operations | Correct the owned source of truth and reachable third-party profiles. |
| Competitors cited from third-party sources | PR, partnerships, or founder/subject expert | Pursue legitimate source inclusion or expert contribution. |
| Negative or weak sentiment | Customer success, product, or comms | Fix the underlying issue and update the source trail. |
| Technical access issue | Web, SEO, or engineering | Repair robots, noindex, canonical, renderability, sitemap, or structured data issues. |
| Unclear prompt set | Growth or analytics owner | Rewrite, retire, or regroup prompts before reporting movement. |
Do not let "marketing" own the whole audit. A real owner is a named person or function with the authority to change the source that caused the gap.
Where ReachLLM fits
ReachLLM is built for teams that want the audit connected to execution.
The platform tracks prompt portfolios across enabled AI systems, preserves raw answers, compares competitors, and measures Visibility Score, Share of Voice, Average Rank, citation rate, sentiment, source data, and history. It also supports GEO audits, source intelligence, brand-fact review, content updates, website changes, structured data, llms.txt, PR outreach, integrations, governed approvals, agent workflows, and managed execution.
Use ReachLLM when:
| Audit need | How ReachLLM helps |
|---|---|
| Leadership wants evidence behind the number. | Raw answers, citations, competitors, sentiment, and run history stay connected to each metric. |
| The team does not know why competitors win. | Source intelligence separates owned pages, third-party sources, directories, and competitor evidence. |
| Facts are wrong in AI answers. | Brand-fact review connects claims, corrections, owners, and source-of-truth updates. |
| Findings do not become shipped work. | GEO audits, content updates, website fixes, schema, llms.txt, PR, and managed execution can sit in the same workflow. |
| Reporting needs before-and-after proof. | The same prompt group can be re-run after the fix ships. |
If a team only needs a one-time manual snapshot, a spreadsheet can work. If the team needs repeatable audit, diagnosis, execution, and proof, use a system that keeps those steps connected.
A 60-minute AI visibility audit meeting
Use this agenda when the team already has some data and needs to turn it into action.
| Minute | Work |
|---|---|
| 0-5 | Confirm the business objective, market, language, and decision owner. |
| 5-10 | Review the prompt groups and remove anything stale, duplicate, or ownerless. |
| 10-20 | Read the highest-intent raw answers, not just the dashboard. |
| 20-30 | Separate branded accuracy, category discovery, comparison, trust, and capability gaps. |
| 30-40 | Open the cited sources and classify source types. |
| 40-50 | Route the top gaps to owned-page, technical, source, PR, profile, product-fact, or positioning fixes. |
| 50-55 | Assign one owner and one next action per priority gap. |
| 55-60 | Set the re-measurement date and exact prompt group. |
The output should fit on one page:
| Field | Example |
|---|---|
| Scope | 40 US English prompts across ChatGPT Search, Google AI Overviews, Gemini, and Perplexity. |
| Main gap | Category prompts cite competitors from two roundups and one directory while the brand is absent. |
| Source diagnosis | Owned category page is vague; third-party roundups omit the brand; one profile is stale. |
| Fix | Rewrite the category page, update the directory profile, and pitch one legitimate cited source with proof. |
| Owner | Product marketing owns the page; partnerships owns profile/source work. |
| Re-check | Same category prompt group two weeks after publishing and source updates. |
That is more useful than a 40-slide audit that nobody owns.
What not to do
- Do not treat one AI answer as a trend.
- Do not merge branded and unbranded prompts into one comfort score.
- Do not count a mention as a citation.
- Do not assume a citation is a recommendation.
- Do not publish a new page for every missing prompt.
- Do not copy competitor articles or source pages that currently get cited.
- Do not claim schema,
llms.txt, crawler access, or one page update guarantees AI placement. - Do not report movement without checking whether the prompt set, competitor set, market, or platform changed.
The best audit makes the next action smaller, clearer, and easier to verify.
FAQ
What is an AI visibility audit?
An AI visibility audit reviews how a brand appears across AI answer surfaces. It checks prompt scope, brand mentions, rank, competitors, citations, source URLs, sentiment, factual accuracy, technical eligibility, and the fixes needed to improve future answers.
How is an AI visibility audit different from a GEO audit?
A GEO audit often includes website readiness, content structure, entity clarity, and source strategy. An AI visibility audit starts from live answer evidence: prompts, raw answers, mentions, citations, sources, sentiment, and competitors. The strongest workflow uses both.
How often should teams run an AI visibility audit?
Run a full audit monthly or quarterly, depending on team capacity. During launches, pricing changes, rebrands, or major source updates, re-run the affected prompt groups sooner so the team can validate before-and-after movement.
Should every audit gap become a new article?
No. A gap may require an existing page update, technical repair, profile correction, PR outreach, source inclusion, product-fact cleanup, competitor-set change, or prompt rewrite. Publish a new article only when it adds original value for a real buyer question.
Does technical SEO guarantee AI citations?
No. Crawlability, indexability, schema, sitemaps, and crawler access support eligibility and clarity, but they do not guarantee citations or recommendations. Treat technical work as audit hygiene, then inspect answer and source evidence.
How does ReachLLM help with AI visibility audits?
ReachLLM connects AI visibility audits to execution. Teams can track prompts, inspect raw answers, compare competitors, review citations and source intelligence, measure sentiment and visibility, run GEO audits, update content and pages, manage brand facts, coordinate approvals, and re-measure after fixes ship.
Sources reviewed
- Ahrefs, "AI Visibility Audit: How to Measure Your Brand's Presence in AI Search": https://ahrefs.com/blog/ai-visibility-audit/
- Google Search Central, "AI features and your website": https://developers.google.com/search/docs/appearance/ai-features
- Google Search Central, "Creating helpful, reliable, people-first content": https://developers.google.com/search/docs/fundamentals/creating-helpful-content
- Google Search Central, "Robots meta tag, data-nosnippet, and X-Robots-Tag specifications": https://developers.google.com/search/docs/crawling-indexing/robots-meta-tag
- OpenAI, "Overview of OpenAI crawlers": https://developers.openai.com/api/docs/bots
- OpenAI Help Center, "Searching the web with ChatGPT": https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt
- ReachLLM Platform page: https://www.reachllm.com/platform
- ReachLLM Complete Platform Capabilities: https://www.reachllm.com/platform/capabilities