The 2026 AI Search Visibility Audit Template

Here’s the 2026 reality check: only 6.8% of U.S. ChatGPT desktop answers included a link to an external source as of May 2026 — up more than fivefold from about 1% a year earlier, but still meaning 93 out of every 100 ChatGPT answers send users nowhere (Similarweb’s 2026 Generative AI Landscape report, via Search Engine Journal). Meanwhile, Google’s AI Overviews now reach over 2 billion monthly users across more than 200 countries (Ahrefs).

The window where AI search visibility is a nice-to-have is closed. What’s open is a different question: which surfaces matter for your category, and whether your brand shows up in the answer. That’s what this template audits.

TL;DR

• Being recommended is not the same as being cited. Across Google AI Mode, ChatGPT, and Perplexity, only 2.8% of the 1,851 sources cited were brand-owned pages, while 59% of citations went to third-party sites — even when the AI recommended the brand by name (Shero Commerce, Aug 2026). Your audit must measure both the recommendation and the citation.
• 89.3% of estimated AI search demand sits inside categories with no clear owner yet. Kevin Indig’s analysis of 1,094 U.S. categories (five prompts each, 50,000+ brands, 600,000+ citations, January–June 2026) found only 15.2% of categories had a clear owner (Search Engine Journal). The audit’s real prize is category ownership, not a mention here and there.
• Brand mentions predict AI visibility far better than authority metrics. Across 75,000 brands, Ahrefs measured YouTube mentions at a correlation of ~0.737 and branded web mentions at 0.656–0.709, versus Domain Rating at just 0.266–0.326 (Ahrefs).
• Ranking in the top 10 no longer means being cited. Only ~38% of URLs cited in AI Overviews also ranked in the top 10 for the same query, down from ~76% a year earlier (Ahrefs, 863K SERPs and 4M citations). Semrush’s 200,000-keyword study found the top-ranked result appeared in just 46% of desktop AI Overviews (Semrush).
• AI Overviews are expensive when they appear: Ahrefs measured a 58% cut to position-1 CTR in 2026, up from 34.5% in 2025 — corroborated by Seer (65.2%), Kevin Indig (>50%), and Authoritas (47.5%) (Ahrefs).
• Don’t spend your 90 days on the shortcut stack. 97% of published llms.txt files received zero requests in May 2026 (Ahrefs, 137,210 domains), and adding JSON-LD schema to 1,885 pages produced no citation uplift versus 4,000 control pages (Ahrefs). Technical hygiene matters; it just isn’t the lever.
• The tools are cheap enough to start today. Dedicated AI visibility tracking starts around $29/month (Otterly AI), with Mid-market platforms from $95–$99/month (Peec AI, Profound, Semrush AI Visibility) (Semrush tool comparison).

What this guide covers

1. What an AI Search Visibility Audit Measures in 2026
2. Step 1: Build Your Money Prompt Set
3. Step 2: Run the Manual Baseline Sweep
4. Step 3: Score Brand Appearances with the 5-Check Rubric
5. Step 4: Audit Citations and Sources
6. Step 5: Entity Consistency and Off-Site Presence
7. Step 6: Technical Access and AI Crawler Audit
8. Step 7: Competitive Gap and Category Ownership Analysis
9. Step 8: Measure Impact — AI Referrals and KPIs
10. The AI Visibility Toolstack, Compared
11. Scoring and Prioritization Framework
12. The 30/60/90-Day Audit-to-Roadmap Plan
13. Frequently Asked Questions
14. Sources and References

What an AI Search Visibility Audit Measures in 2026

An AI search visibility audit is a structured assessment of how your brand shows up in AI-generated answers: whether you’re mentioned, whether you’re cited, whether the description is accurate, and whether someone else owns the answer for the questions that matter to your business.

The critical distinction in 2026: mentions and citations are only weakly related. In Kevin Indig’s 600,000-citation dataset, citation volume and brand mentions correlated at just -0.229, and the most-frequently-cited domain matched the most-mentioned brand only 20.8% of the time — while the most-mentioned brand was cited at least once 69.9% of the time (Search Engine Journal). A brand can be named without being trusted, and trusted without being named. Audit both, separately.

Why the audit is urgent now: Google says AI Overviews reach over 2 billion monthly users (Ahrefs), Google AI Mode passed a billion monthly users in 2026 (Search Engine Journal), and standalone chatbots now draw 655 million monthly unique visitors versus classic search’s 3.3 billion — with 461 million of ChatGPT’s 494 million users also using Google (Similarweb via Search Engine Journal).

Which platforms to audit

Prioritize by where your audience is, not by hype. Audience preference splits sharply by generation: ChatGPT leads every age group (44.4% of Gen Z AI users to 24.5% of Boomers), Gemini leads Gen X at 22.8%, Claude’s Gen Z preference is 10.6% versus just 1.4% among Boomers (a ~7.6x gap), and Copilot skews opposite — 12.3% among Boomers but 4.4% among Gen Z (YouGov U.S. AI brand rankings, via Search Engine Journal).

Platform	Why audit it	What to measure	Data you’ll capture
Google AI Overviews	Largest reach: 2B+ monthly users; appears on 68% of local business-type queries (vs 39% local packs, per Whitespark) and >80% of “price/cost/buy” queries	Presence share, citation share, URL overlap with your rankings, sentiment	Prompt sweep + Ahrefs Brand Radar-style mention share + GSC impression data
Google AI Mode	1B+ monthly users; queries run ~3x longer than traditional search; ads are live and cannot be blocked by advertisers	Sidebar domain inclusion, citation of your pages, branded query lift	Prompt sweep + Merchant Center AI Performance Insights (limited U.S. pilot)
ChatGPT	Sends 85.8% of all AI-platform referral traffic; links out in only 6.8% of U.S. desktop answers	Citation rate, brand mention, description accuracy, recommendation strength	Similarweb/GAA tracking from chat.openai.com
Gemini	Strongest with Gen X; citations skew to community and reference surfaces	Mention + citation, position in comparison answers	Prompt sweep
Perplexity	Citation-forward publisher economics (Comet Plus pays publishers 80% of revenue from a $42.5M pool)	Citation share, source ranking; typical answers cite 4–8 sources	Prompt sweep + Perplexity partner analytics
Copilot	Strongest with older users; enterprise buyers lean in	Mention + citation	Prompt sweep

Sources for the platform table: Search Engine Journal (Whitespark, Averi.ai, Omniscient), Search Engine Journal, Semrush, LLM Pulse.

“AI search isn’t unknowable, but it’s definitely allergic to hacks.” — Ahrefs, 15 million data points across 50+ AI search studies, 2026

Step 1: Build Your Money Prompt Set

The measurement unit of an AI visibility audit is the prompt, not the keyword. Google’s own guidance to SEOs says you don’t need special files or markup — but you do need your content to answer the questions people ask in these tools (Google, via Search Engine Journal).

The evidence you need a prompt set, not a keyword list: Go Fish Digital ran 73 unique prompts (1,554 completed responses) across ChatGPT, Google AI Overviews, and Perplexity in July 2026. When the prompt was category-led (“best GEO tool”), their mention rate was 23.6%; problem-led prompts (“no industry label”) dropped to 7.6%; full buyer-situation prompts fell to 3.6% — a 6.6x gap driven purely by prompt wording (Go Fish Digital). Your prompt choices bias your score more than your content does.

Build the set — checklist:

• List 25-30 prompts across three types: branded (“what does [brand] do”, “is [brand] worth it”), category (“best [category] for [use case]”, “top [category] alternatives”), and competitor (“[competitor] vs [your brand]”)
• Mirror Kevin Indig’s five proven prompt types: definition, comparison, alternatives list, use case, and buying question (Search Engine Journal)
• Note the buyer stage: awareness-stage (research) and decision-stage (“should I buy”) prompts both earn a place
• Score each prompt 6-22 on the Semrush model: buyer stage (x3 weight), topic relevance (x3), competitor presence (x2). Band 18-22 = core target; 12-17 = track if there’s room; below 12 = drop (Semrush)
• Cap the set near 25 prompts to match the allowance on entry-level AI visibility plans — cheap validation that your set is monitorable long-term
• Cluster prompts by intent and report at category level, not per prompt: AI systems expand one query into related variants, so one good prompt covers many variations (Semrush)

Validation rule: keep a prompt if (a) your brand is cited, (b) a competitor is cited and you’re absent, or (c) it’s a high-priority buyer situation with no brand presence. Retire prompts with no citation or mention signal for 60+ days, and review the whole set quarterly.

Step 2: Run the Manual Baseline Sweep

Before you buy or configure anything, run the manual sweep. It’s the fastest way to detect the gap, and it produces the baseline your tools will later ratify. Go Fish Digital’s step-by-step GEO audit is the cleanest published version of this workflow (Go Fish Digital).

Sweep procedure:

• Clean the session. Use temporary chat, private mode, or a new account in ChatGPT, Gemini, Claude, Perplexity, or Copilot — personalization from prior conversations skews results
• Ask the broad category question: “What are the best [products or services in your category]?” Record whether your brand appears
• Ask the buyer-constraint question: “I’m a [type of customer] dealing with [specific problem]. What should I look at?” — the recommended list often shifts when a real constraint is introduced
• Prompt the head-to-head: “Compare [brand] and [competitor]. Which would you recommend for [specific need]?”
• Ask the identity questions: “What does [brand] do?”, “What is [brand] known for?”, “What do customers say about [brand]?” — then document what AI says versus what you know is true. “That gap between what AI says and what you know to be true is your baseline” (Go Fish Digital, Aug 2026)
• Repeat in at least two assistants, more than once per prompt, and log every brand named, recommended, and linked
• Run prompts with and without industry vocabulary — the 6.6x wording gap means vocab-free, realistic buyer prompts will show your true baseline (Go Fish Digital)

Step 3: Score Brand Appearances with the 5-Check Rubric

For every prompt-platform combination, score the response against five checks. Go Fish Digital publishes this model as one-, two-, and three-check scoring with clear verdict bands (Go Fish Digital).

Checks passed	Verdict	Action
5 of 5	No gap	No action; move to monitoring
4 of 5	Small gap	Quick pass — tighten one weak element
2-3 of 5	Real gap	Full remediation (Steps 4-7)
0-1 of 5	Effectively invisible	Prioritize; treat as zero baseline

The five checks:

1. Appearance — is your brand in the answer at all? (Averi.ai found Reddit is the most-cited source in AI Overviews at 21% — competing for a slot against community platforms, not just your rivals.)
2. Recommendation strength — are you actually picked for the constraint, or one name among many?
3. Link/citation — does the answer include a clickable link to your site? Remember the Shero finding: when brands were recommended, their own page was cited only 31% of the time (Search Engine Journal).
4. Head-to-head win — the comparison prompt answers who’s recommended, not just who’s mentioned.
5. Description accuracy — products, audience, pricing, positioning correct? Document what’s stale or wrong.

That’s the brand-level needle. For the category-level needle, add a diversity check: in Semrush’s five-vertical AI Visibility Study, consumer electronics responses named only ~1-2 brands per query (1.22 on ChatGPT, 1.36 on AI Mode) while business services named ~5 (4.72 / 4.60) — some categories are one-slot games, others are five-slot games. You need fewer wins in the former (Semrush).

Step 4: Audit Citations and Sources

Citations are where AI’s trust lives, and where your brand’s absence is most fixable. Four datasets define the citation landscape in August 2026:

• Third-party ecosystems dominate. Ahrefs’ July 2026 AI Overviews study (3M+ U.S. queries) put YouTube at 21.1% of top-source citations, Reddit at 18.5%, then Facebook 10.7%, Google 7.1%, Instagram 5.8%, Quora 5.0%, Wikipedia 4.8%, Amazon 3.5%, TikTok 3.2%, and Walmart 1.0% — the top ten together absorb roughly 80% of cited-source attention (Ahrefs). Omniscient Digital’s 23,000+ citation analysis found only about 23% of citations in branded queries go to owned content — 77% is off-page (Search Engine Journal).
• Editorial coverage converts to mentions. A Leoprd study found 61.9% of citations that mention a brand come from editorial coverage, awards, and reviews — not from content the brand publishes itself (Semrush).
• Be first, be fresh. Kevin Indig’s analysis of 1.2 million ChatGPT responses found 44.2% of citations come from the first 30% of the page — answers buried in prose get lost (Search Engine Journal). Ahrefs’ analysis of 17 million AI citations found AI-cited content is 25.7% fresher than traditionally ranked content.
• Reward structure is changing. Perplexity’s Comet Plus pays publishers an 80% share of subscription revenue from a $42.5 million pool across human visits, search citations, and agent actions — with typical answers citing four to eight sources and multiple citations multiplying payouts (LLM Pulse).

Citation audit checklist:

• For each prompt, record which domains get cited and in what order — not just whether you appear
• Flag competitor citations from review platforms and marketplaces (Trustpilot climbed 11 places in Ahrefs’ monthly snapshot to 0.4% mention share) — a signal your own profiles there are stale (Ahrefs)
• Identify your most-cited page and note where it sits: 65% of URLs ChatGPT cites sit two or three folders deep (Aleyda Solis, Similarweb 2026 report) — depth is normal, thickness is not
• Check freshness: update date-affected pages and show visible last-updated dates
• Put the direct answer in the first sentences of each section — the first 30% of a page decides most citation wins
• Compare the cited competitor’s page against yours on five axes: directness of answer, evidence strength, standalone comprehensibility, authority signals (bylines, credentials, publisher), and quotability (Go Fish Digital)

Step 5: Entity Consistency and Off-Site Presence

AI has no master database. It assembles a brand picture from whatever content is “most consistent and credible” across sources — your site, directories, Yelp, Reddit, Wikipedia, press. “When those sources agree, AI answers confidently. When they conflict or go stale, AI fills gaps on its own” (Go Fish Digital, Aug 24, 2026). The same article’s OREO test shows even deeply branded companies can be “a ghost in AI search”: a natural-language query in Google AI Mode returned generic cookies, no OREO.

Entity consistency checklist:

• Standardize positioning language exactly — same description on website, Google Business Profile, LinkedIn, Crunchbase, and key directories, with no paraphrasing
• Publish explicit entity content: pages that state directly who you are, what you do, who you serve, and your differentiators in quotable language
• Audit off-site sources in priority order: Wikipedia/Crunchbase/GBP/major directories carry more weight than smaller listings
• Check brand-name ambiguity: common words, lookalike competitors, and one-letter-off names create entity confusion
• Clean up legacy narratives: rebrands, discontinued products, and outdated pricing on indexed pages “quietly working against” you — and remember affiliate sites are heavily crawled
• Use structured data for long-term entity associations (Organization, Person, sameAs to Wikipedia/Wikidata/Crunchbase) — the one place Ahrefs’ schema experiment showed real value, which is long-term entity clarity, not citation uplift (Search Engine Journal)

Third-party presence checklist:

• Earn fresh mentions: digital PR, reviews, analyst coverage, community answers. That’s the highest-leverage category — 61.9% of brand-mention citations come from editorial-style sources
• Build a YouTube presence: creator partnerships, reviews, tutorials. YouTube mentions carry the strongest measured correlation with AI visibility (~0.737 across 75,000 brands) (Search Engine Journal)
• Listen on Reddit, don’t pitch. Communities are skeptical of brand presence; observe recurring questions and misperceptions, then close the gaps through earned coverage elsewhere (Go Fish Digital)
• Consider Google’s Preferred Sources — 600,000+ sources have been selected, up from 345,000 in May 2026, and preferred sources are increasingly infused into AI Overviews and AI Mode (Search Engine Journal, Aug 25, 2026)
• Document your 12-18 month horizon: agencies working across local verticals estimate that’s the window before citation authority settles (Search Engine Journal)

“AI doesn’t experience campaigns. It reads online text.” — Go Fish Digital, “Why AI Gets Your Brand Wrong,” August 2026

Step 6: Technical Access and AI Crawler Audit

This step is faster than any other and eliminates the failure modes that make everything else pointless.

• Check robots.txt and access rules: confirm GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, and Google-Extended aren’t blocked on commercial content. Note the trade-offs: blocking some training crawlers while allowing retrieval bots is a defensible decision, but make it explicit, not accidental
• Audit AI crawler activity in log files: agentic traffic crossed 50% of all internet traffic for the first time in June 2026 (Cloudflare, via Ahrefs). Your server logs distinguish crawler noise from retrieval-bot visits — know which AI engines are actually reading you
• Fix weak internal context: Semrush’s audit guidance flags links without descriptive anchor text and pages with only one incoming internal link as the top structural reasons citations come from a small set of pages (Semrush)
• Confirm pages are citable: they must load normally, not sit behind login, and offer a real HTML version rather than a PDF-only asset (Go Fish Digital)
• Steer the Claude-Code-adjacent caveat: AI agent traffic is the largest AI bucket reading site files — but 97% of llms.txt files get zero requests, and 77% of the remaining fetches aren’t AI bots at all. Skip the llms.txt gold rush: make content easy to crawl, parse, and cite instead. John Mueller has called it “a temporary crutch” (Ahrefs)
• Don’t expect schema to move citations. Ahrefs tracked 1,885 pages that added JSON-LD against 4,000 controls over 30 days: no meaningful uplift on AI Mode or ChatGPT, and a small (-4.6%) decline on AI Overviews. Treat markup as hygiene with entity value, not a citation lever (Search Engine Journal)

One structural reality check for content teams: AI Overviews and AI Mode share 86% semantic similarity but only 13.7% citation overlap (Ahrefs, via Go Fish Digital). Winning one surface does not buy you the other — audit them separately.

Step 7: Competitive Gap and Category Ownership Analysis

This is the step that turns a visibility audit into a strategy. The headline: in June 2026, only 15.2% of categories had a clear owner, and 89.3% of estimated AI search demand sits in categories with no clear owner yet — the higher-volume half of categories had an owner rate of just 11.3%. Owners held first place across 90.4% of month-over-month comparisons, and the categories that flipped had a median lead of just 1.3 percentage points going into the switch (stable ones: 2.9) (Search Engine Journal).

Gap analysis checklist:

• Classify your categories: clear owner (one brand in 4 of 5 prompts, beating runners-up by 5+ points), emerging leader, or unsettled (no brand leads 3 of 5 prompts)
• Treat any lead under ~3 points as contested — that’s where ownership flips month to month
• Run Semrush-style topic gap analysis: find topics where AI answers exist and competitors are mentioned but your brand is absent; expand each topic to the specific missed prompts, then assign them as content briefs (Semrush)
• Watch the “most mentioned but never cited” pattern: in Semrush’s study, Zapier was the #1 cited source in digital technology but only #44 in brand mentions, and only 6-27% of most-mentioned brands were also top cited sources — being talked about doesn’t equal being trusted (Semrush)
• Benchmark share of voice within your category — Semrush’s index shows what ownership looks like: Samsung 58.08% vs Apple 48.84% in consumer electronics; Google 23.22%, Zoho 16.70%, HubSpot 15.40% in business services (Semrush)
• Prioritize citation-building on comparison and educational content — “educational and comparison pages earn more citations than product or category pages” (Semrush)

One honest caveat to measure against: in Jan-Willem Bobbink’s GEO tactics test, an untouched page was cited 13.3% of the time on GPT-4o-mini while the GEO-treated page got 10.9–12.2% — “the tactics everyone is selling lost to doing nothing at all” (Ahrefs). And Wil Reynolds’ personal-site experiment showed a 1,900% month-over-month jump in ChatGPT citations to a single page with little-to-no business impact. Citations are a means, not the end.

Step 8: Measure Impact — AI Referrals and KPIs

The last step is making the audit renewable. Three measurement traps are documented in 2026 data:

1. GA4 misattributes AI traffic. “Many AI clicks show up as ‘direct’ traffic in Google Analytics 4 because AI platforms often don’t pass referrer information,” meaning AI traffic is likely underreported (Semrush). If you’re not building AI-referral segments by domain (chat.openai.com, copilot.microsoft.com, perplexity.ai, gemini.google.com), your dashboard is lying to you.
2. Search Console is partial. Clicks and impressions from AI Mode now count inside the standard Performance report’s “Web” search type — but there’s no dedicated AI tab, and Google’s Generative AI report in GSC provides impression data only (no clicks, no queries). Brodie Clark called it “the most requested report in Google Search Console ever” — and it still under-delivers (Ahrefs). Bing Webmaster Tools, by contrast, exposes AI citation performance data.
3. Cited pages and clicked pages diverge. 65% of URLs ChatGPT cites sit deep in the site, while 58.8% of referral traffic lands on the homepage (Aleyda Solis, Similarweb 2026 report) — so a citation-only dashboard misses where humans arrive, and a traffic-only dashboard misses the trust you’re building (Search Engine Journal).

KPI dashboard (7 metrics, per LLM Pulse’s GEO framework (LLM Pulse)):

Metric	What it answers	Where to capture
Brand mentions	Is the brand name present at all? (bare URLs don’t count)	Prompt sweeps + tracking tools
Brand visibility	How consistent is presence across the prompt set?	Tracking tool dashboards
Share of voice	Am I mentioned more than competitors?	Tracking tools
Sentiment	Is the framing favorable, neutral, or warned?	Tracking tools; for the audit, manual review
Citations & sources	Which platforms link you — and which link competitors?	Prompt sweeps; Ahrefs Brand Radar-style studies
AI referral traffic	What lands on site, and where?	GA4/Looker Studio segments
Revenue & influence	Do AI-exposed users convert?	GA4 assisted conversions, form/lead dashboards

Business impact framing for leadership:

• AI search visitors convert 4.4x more likely than traditional search visitors, per Semrush research, which also predicts AI search traffic could exceed traditional search traffic by 2028 (Semrush)
• In a law-firm case study, the three most-clicked AI referral pages were the homepage, contact form, and confidential submission page — “AI referrals are not just noise. They are a new form of high-intent traffic that bypasses traditional funnels” (Go Fish Digital, Aug 7, 2026)
• Keep expectations honest: Google still sends 190x more traffic than ChatGPT across 76,000 sites (ChatGPT = 0.21% of site traffic), and ChatGPT sees only ~12% of Google’s search volume for traditionally Googled queries (Ahrefs, via Search Engine Journal). The 190x chart is why you keep doing SEO; the 6.8% citation rate is why you audit AI visibility
• Monitor the “crocodile mouth” pattern: pages where impressions hold steady and clicks fall away — that’s an AI Overview absorbing clicks, not a ranking problem (Ahrefs)

The AI Visibility Toolstack, Compared

Prices verified August 2026 via Semrush’s tool comparison:

Tool	Entry price	Free option	Best for your audit
Otterly AI	$29/mo (Lite)	Free trial	Lowest-cost monitoring; GEO audits covering 25+ on-page factors; keyword-to-prompt conversion
HubSpot AEO	$50/mo ($45 annual)	Free trial, no card	Cheapest dedicated AEO entry; 25 tracking prompts; prioritized next steps
Peec AI	$95/mo (Brands Starter: 50 prompts, 3 AI models)	Trial on select plans	Mid-market; alerts on sudden visibility gains/losses; 80+ countries
Profound	$99/mo	None	Analyst-grade raw data; 1.5B+ real user prompts; agent analytics for AI crawler behavior
Semrush AI Visibility	$99/mo per domain	7-day trial	Combined SEO + AI visibility; 289M+ prompt database; prompt research/tracking; includes AI Overviews and AI Mode coverage
Writesonic	$99/mo	None	Visibility tracking plus content execution; 2B+ AI conversations across 10 platforms
Athena	Free tier ($25 credits); $295/mo Starter	Yes — free Essential	Keyword-level AI visibility tied to GA4 + Search Console traffic
Botify	Custom	No	Enterprise: 50M-URL crawls, log-file AI bot analysis, automated fix deployment

Tool selection rules of thumb: start manual (Step 2-3) before buying — a prompt set validated by hand will work in any tool. Buy tracking when your prompt set is stable and you need time-series data. The agencies audited here used Semrush, Peec AI, Otterly AI, and Loamly loosely interchangeably for LLM mention-tracking (Go Fish Digital).

Scoring and Prioritization Framework

Aggregate the audit into a score your team can act on. Use two thresholds:

A. Category readiness score (0-100) — weight by revenue-relevance, not by convenience:

Component	Weight	What pushes the score
Mention rate on core prompt set	25%	Records from Step 3 rubric
Citation presence	25%	Own-page citations vs competitor citations
Entity consistency	15%	Consistency across GBP, Crunchbase, LinkedIn, site
Off-site presence	15%	Editorial/community/YouTube mention coverage
Technical access	10%	Crawler access, indexability, citable pages
Measurement capability	10%	Can you re-run and report in <30 days?

Score bands: below 60 — do Step 1-6 before any content expansion; 60-79 — targeted gaps remain, usually off-site presence and comparison-content coverage; 80-100 — shift to category-ownership push and monthly monitoring.

B. Prompt-level priority (from Step 1’s 6-22 scoring): attack 18-22 first, protect 12-17 with quarterly content refreshes, and drop anything below 12.

Sequencing principles from the data:

• Fix technical access first — it’s the only failure mode that makes every other investment worthless, and it’s typically a week of work (Semrush)
• Spend 3x the effort on off-site presence and third-party citation surface versus on-site rewriting — that’s where 77% of branded-query citations live
• Don’t scale content volume to chase coverage: Ahrefs found content volume correlates only ~0.194 with AI visibility while brand mentions dominate (Search Engine Journal)
• Prepare for the ads layer in your owned surfaces: AI Mode text ads appeared on 29% of tested commercial keywords (SE Ranking), the advertised domain matched a cited source only 11% of the time, and advertisers cannot block AI Mode placements — the organic game and paid game are converging (Search Engine Journal, Search Engine Journal)

The 30/60/90-Day Audit-to-Roadmap Plan

Days 0-30: Baseline and remove blockers

• Build and validate the 25-prompt money set (Step 1)
• Run the manual sweep in at least two assistants; complete the 5-check scoring
• Fix robots.txt/access issues; audit AI crawler logs; fix weak internal linking (Step 6)
• Set GA4 AI-referral segments by platform domain; established baseline mention and citation rates
• Select and deploy one tracking tool if the set is stable

Days 31-60: Citation readiness

• Restructure money pages so answers live in the first 30% of each section, with visible update dates
• Build educational and comparison content identified by the gap analysis (educational/comparison pages earn the most citations)
• Clean entity consistency: canonical descriptions, GBP, Crunchbase, LinkedIn; publish explicit entity pages
• Launch a digital PR + creator (YouTube) push; start the 12-18 month off-site flywheel
• Re-run the sweep weekly for one monitorable category; note prompt-wording effects (remember the 6.6x gap)

Days 61-90: Ownership push and measurement

• Re-run category classification monthly; treat <3-point leads as contested and target the highest-volume unsettled categories
• Benchmark share of voice against competitors; verify sentiment and description accuracy of your own brand
• Publish the first before/after report: mention rate, citation rate, SOV, AI referral sessions, and AI-assisted conversions
• Refresh the prompt set quarterly (add launch/positioning prompts, retire 60+ day dead prompts)
• Reassess platform priority with new audience data — preference splits by generation and shifts fast (Claude’s Gen Z preference alone nearly doubled in Q2 2026)

Frequently Asked Questions

What’s the difference between an AI visibility audit and a traditional SEO audit? Traditional audits measure outcomes — rankings, clicks, sessions — in search engines. AI visibility audits measure readiness for selection: whether your content gets retrieved, cited, and recommended inside AI-generated answers across ChatGPT, AI Overviews, AI Mode, Gemini, Perplexity, and Copilot. The gap between the two is now measurable: only ~38% of URLs cited in AI Overviews also ranked in the top 10 for the same query, down from ~76% a year earlier (Search Engine Journal).

Why do I need to audit mentions and citations separately? They’re only weakly related. Citation volume and brand mentions correlated at -0.229 across 600,000+ citations, and the most-mentioned brand was cited at least once 69.9% of the time — meaning mentions overstate trust, and citation rankings reveal who AI leans on (Search Engine Journal). In ecommerce specifically, brands were recommended while third-party sites got the citations 59% of the time (Search Engine Journal).

Which platforms should I audit first? Start where your buyers are, then expand. ChatGPT is the biggest referral source (85.8% of AI-platform referral traffic) but cites sparingly (6.8% of U.S. desktop answers linked out in May 2026); AI Overviews have the largest reach (2B+ monthly users) and show up on 68% of local business-type queries; AI Mode passed a billion monthly users and is where commercial queries go long. Generational data helps: ChatGPT leads every age group, Gemini leads Gen X, Copilot skews older, Claude skews young (Search Engine Journal).

Does schema markup help get cited? Not measurably. Ahrefs’ controlled test of 1,885 schema-added pages versus 4,000 matched controls found no meaningful citation uplift on AI Mode or ChatGPT and a small decline on AI Overviews. Its value is long-term entity clarity (Organization, Person, sameAs to Wikipedia/Wikidata/Crunchbase) and rich results consistency — hygiene, not a lever (Search Engine Journal).

Should I publish an llms.txt file? Only if you want to, and don’t expect anything from it. Ahrefs found 28% of 137,210 domains publish one, but 97% of published files got zero requests in May 2026 — and 77% of the requests that did happen weren’t AI bots. Zero AI bots requested a non-existent file, and generative-engine optimization tools were a top fetching category; the file signals the industry studying itself more than agents reading you (Ahrefs).

How do I measure AI traffic when GA4 shows it as direct traffic? Many AI referrals arrive without referrer data and get bucketed as “direct” in GA4, so actual AI traffic is likely underreported. Set up segments by known AI referral domains (chat.openai.com, copilot.microsoft.com, perplexity.ai, gemini.google.com, plus you.com and iask.ai) and build a dedicated AI-referral view in Looker Studio comparing AI versus organic, paid, and direct performance on engagement, form submissions, and conversions (Semrush, Go Fish Digital).

How long until AI visibility moves? Expect the baseline sweep and technical fixes within weeks, citation gains within a quarter, and category ownership over 12-18 months — the window practitioners estimate before citation authority settles in local and competitive categories (Search Engine Journal). Category leaders are sticky once established (owners held first place in 90.4% of month-over-month comparisons), which is exactly why the unsettled 89.3% of demand matters now.

Is it worth it if Google still dwarfs ChatGPT? Yes, but sequence it. Google sends roughly 190x more traffic to sites than ChatGPT does (0.21% of site traffic versus ~40%), per Ahrefs’ 76,000-site panel — so classic SEO stays the volume engine (Search Engine Journal). But AI-referred users convert at 4.4x the rate of traditional organic visitors and predominantly arrive at high-intent pages like contact and submission forms (Semrush, Go Fish Digital). The audit positions you where the margins are, without abandoning the volume.