Key takeaways
- Ranking on page one no longer guarantees a citation. AI Overviews cite sources that overlap with the organic top 10 only 16.7% of the time, so a content audit built for AI search has to check different signals than a classic SEO audit.
- Freshness matters more than most teams assume. AirOps' own research on 4,000+ cited pages found more than 70% were updated in the last 12 months, and pages untouched for over a year are twice as likely to lose citations to a competitor.
- ChatGPT only cites about five sources per response, roughly half of what Google AI Overviews and Perplexity cite, which means every content gap you find is fighting for a smaller number of slots.
- Run the audit in layers: technical crawl access, content structure and extractability, authority/E-E-A-T signals, schema, freshness, then measurement. AirOps organizes this as a 48-factor checklist, which is a useful skeleton even if you're not using their tool.
- A one-time audit goes stale fast. Citation share is volatile enough that a quarterly cadence, not an annual one, is what actually catches the gaps before they cost you visibility.
Why an AirOps content audit needs its own playbook
If your team has been producing content through AirOps, running briefs through Quill, or publishing pages via its content workflows, you're probably assuming the output is already AI-search-ready. It might not be. AirOps is built to help teams produce and refresh content at scale, but scale doesn't automatically equal visibility. A page can be grammatically clean, keyword-optimized, and published on schedule, and still get skipped by ChatGPT or Google AI Overviews because of something as mundane as a buried answer three paragraphs deep, or a missing "last updated" date.
AirOps' own data backs this up. According to AirOps research, only 30% of brands stay visible from one AI answer to the next, and just 20% remain visible across five consecutive runs. That's not a content quality problem exactly, it's a volatility problem, and volatility is exactly what an audit is designed to catch before it becomes a pattern.
What "AI search visibility" actually measures
Before auditing anything, get the terminology straight, because teams conflate two different things constantly.
An AI visibility audit is brand-wide. It looks at citation share, mention rate, and share of voice across ChatGPT, Perplexity, Gemini, and Google AI Overviews for your whole domain and your competitors.
An AEO (answer engine optimization) audit is page-level. It checks whether an individual page can be found, extracted, and cited: crawl access, heading structure, direct-answer placement, schema, freshness signals.
Run them in that order. Measure the brand-level gap first so you know which topics and competitors matter, then go page by page on the priority list.
Step 1: Confirm AI crawlers can actually reach your pages
This sounds basic, but it's where a surprising number of audits find their first real problem. Check robots.txt and server logs for GPTBot, PerplexityBot, ClaudeBot, and the newer Meta-WebIndexer crawler, which went from roughly 2% to nearly 38% of all tracked AI crawler requests between mid-July and early August 2026, according to Promptwatch's crawler data. If that crawler shows up heavily in industry-wide logs but never touches your server, something is blocking it, likely a WAF rule or CDN configuration nobody reviewed since last year.
Aleyda Solis made a similar point in an AirOps webinar: a lot of AI visibility problems trace back to basic crawlability gaps, especially on sites leaning heavily on client-side JavaScript that AI crawlers can't render the way Googlebot does. If your AirOps-published pages sit on a JS-heavy CMS, this is worth checking first, because everything downstream in the audit is irrelevant if the crawler never gets past the door.
Step 2: Audit content structure for extractability
Once you know the crawler can reach the page, the next question is whether it can pull a clean answer out of it. AI systems don't read a page the way a human does, skimming for the gist. They extract a passage and either quote it or paraphrase it. That means structure matters almost as much as the facts themselves.
Check each priority page for:
- A direct answer to the implied question within the first sentence or two of the relevant section, not buried after three paragraphs of context.
- Heading hierarchy that maps tightly to the content underneath it. A heading that says "Pricing" followed by three paragraphs about integrations confuses extraction.
- Short, focused sections rather than one long page trying to cover ten sub-questions loosely.
That last point matters more than it looks. Promptwatch's data on sources per response shows ChatGPT typically cites around five sources per web-search response, about half of what Google AI Overviews (roughly ten) and Perplexity (also close to ten) cite. Fewer slots means more competition per query, and retrieval systems favor content that answers one question fully over content that answers five questions partially. If your AirOps briefs are producing broad, multi-topic pages, split them.
It also helps to write section headings that read like the short queries people (and AI fanout systems) actually type. Promptwatch's query fanout data found the average ChatGPT search query dropped from around 117 characters to roughly 53 characters between December 2025 and April 2026, less than half the original length. ChatGPT is searching more like someone typing keywords than asking a full sentence. Headings like "best CRM for small agencies 2026" outperform vague ones like "choosing the right solution for your team."
Step 3: Check freshness against AirOps' own benchmark
This is one of the clearest, most actionable findings in AirOps' research. Their study of over 4,000 pages cited by ChatGPT across 900 high-intent queries in 15 industries found:
| Update recency | Share of cited pages |
|---|---|
| Updated in last 3 months | 35.2% |
| Updated in last 6 months | 53.4% |
| Not updated in over 12 months | 26.2% |
Pages that hadn't been touched in over a year were more than twice as likely to lose their citation to a competitor. For commercial-intent queries the effect is sharper: over 60% of those citations went to content updated within the past six months. Informational content is more forgiving, nearly a third of cited informational pages were over a year old, but if the page you're auditing drives comparisons, pricing, or "best X" searches, treat six months as your real deadline, not twelve.
Pull a list of every AirOps-published page, sort by last-modified date, and flag anything commercial-intent that's crossed the six-month mark. That list becomes your refresh queue.
Step 4: Map content types against what's actually getting cited right now
Citation patterns shift month to month, and auditing against last year's assumptions wastes effort. Promptwatch's citation type data for August 2026 shows product pages leading ChatGPT citations at roughly 28.7%, but how-to content more than doubled over the month, from 4.3% to 9.1%, and documentation nearly doubled too, from 3.3% to 8.2%. Social posts, meanwhile, collapsed from 4.4% to under 1% after August 14, the same day Reddit's share of ChatGPT citations dropped sharply according to Promptwatch's Reddit report.
Translate this into audit action: if your content calendar has been leaning on social-adjacent or listicle content and skipping how-to guides and documentation, that's a real gap right now, not a hunch. Re-check this quarterly, because the mix moves.
Domain authority also plays a role worth checking. Promptwatch's domain rank data found that in August 2026, domains with DR 46-90 held nearly half of all ChatGPT citations, while DR 91-100 domains fell to about 3% by month's end. You don't need top-tier domain authority to get cited; mid-authority sites in the 61-90 range are a reliable target, which is useful context if you're deciding where to place PR or guest content as part of the off-site half of this audit.
Step 5: Don't forget off-site mentions
AirOps research found that roughly 85% of brand mentions in AI search come from third-party pages, not your own domain. This means half of an AI visibility audit has nothing to do with your website at all. Reviews, community threads, partner content, and comparison sites shape how AI systems describe you just as much as your own blog does. If your audit only checks your own URLs, you're missing the majority of the signal.
Tools that can help run this audit
You can do a lot of this manually with server logs, a spreadsheet, and a lot of patience, but it doesn't scale past a handful of pages. A few tools are purpose-built for this work.
AirOps is the obvious one if you're already producing content there. Its Page360 feature unifies citation signals, Google Search Console data, GA4 traffic, and freshness into a page-level view, so you can see which pages are close to a citation but not quite there, which ones get AI traffic without ranking, and which high-citation pages are starting to age out. Its Quill agent runs in the background scanning for content gaps and stale pages continuously rather than waiting for a quarterly audit.
For a dedicated AI visibility and GEO layer that goes beyond AirOps' own tracking, Promptwatch is worth a look, particularly if you want crawler logs (which AirOps doesn't expose in the same depth), Reddit and YouTube citation tracking, and a content gap analysis feature that turns into briefs your team or an AirOps workflow can act on. Promptwatch tracks ChatGPT, Gemini, Claude, Perplexity, Grok, DeepSeek, Copilot, and Google AI Overviews and AI Mode, with prompt volumes and difficulty scoring baked in, so you're not just seeing that a gap exists but how much search demand sits behind it.
How the audit tools compare
| Tool | Crawler logs | Content generation | Third-party/off-site tracking | Best for |
|---|---|---|---|---|
| AirOps | Limited | Yes, via Quill and workflows | Basic mention monitoring | Teams already producing content in AirOps who want visibility tied to the same pipeline |
| Promptwatch | Yes, 400+ crawlers | Yes, Content Agents with CMS publishing | Yes, Reddit, YouTube, offsite mentions | Teams that want the full loop: why you're invisible, where AI finds you, and automated fixes |
| Search Console + manual logs | Partial (Google only) | No | No | Small sites doing a first pass without buying new software |
If you want to browse more options before settling on one, the GEO software directory at bestgeosoftware.com lists a wider set of platforms by feature set and pricing.
Turning audit findings into a content roadmap
Once the audit is done, resist the urge to fix everything at once. Score each finding by two things: how much AI-driven demand exists for the query it affects, and how far the current page is from being citable. A page that's 80% of the way there with one structural fix (move the answer up, add a schema tag) is a faster win than a page that needs a full rewrite. Build a 30-60-90 day roadmap out of that scoring rather than working the list top to bottom by publish date.
The last step is setting the re-audit cadence. Given how fast citation types and domain rank shares moved within a single month in 2026's data, an annual audit is basically useless for catching drift. Quarterly is the minimum; monthly if you're in a competitive commercial category where six-month-old pages are already losing ground.