After running audits across dozens of brands, we see the same mistakes on repeat. Not one or two per brand. Five, six, sometimes all eight. Each one looks fixable in isolation. The compounding effect is what makes AI visibility hard to self-diagnose.
(The industry calls this practice GEO (Generative Engine Optimization), AEO (Answer Engine Optimization), LLM SEO, or AI SEO. Different names, same work.)
This post maps the eight most common patterns and links to the detailed analysis behind each one.
Assuming SEO has it covered
The most frequent assumption. A brand ranks #1 on Google for their target keyword and assumes they are visible in AI search. They are not. 44% of pages ChatGPT cites don't rank in Google's top 20 at all. SEO optimizes for position in a ranked list. AI visibility is about inclusion in a synthesized answer. The signals are different, the retrieval indexes are different, and the outcomes diverge more every quarter.
We wrote a full breakdown of what this gap costs: Is AI search visibility worth the investment.
Checking only one engine
A brand tests ChatGPT, sees their name in the answer, and assumes they are covered. But only 11% of domains are cited by both ChatGPT and Perplexity for the same query. Each engine retrieves from a different index (ChatGPT from Bing, Claude from Brave, Google AI from its own increasingly divergent results). Visibility on one tells you almost nothing about the others.
CiteGap audits test across ChatGPT, Google AI, and Claude independently for exactly this reason. The full analysis of why engines diverge: The 2-7 domain rule and why each engine searches a different web.
Letting content go stale
Content that ranks well in traditional search can hold position for two to three years. AI engines are different. 50% of Perplexity's citations come from content published in the current year. Content older than 90 days is 2x less likely to be cited by ChatGPT. In our audits, we regularly see thinner but fresher competitor pages winning citation slots over comprehensive but stale brand content. The competitor's page is worse on every metric except the date.
The full data on why recency has become a primary citation filter: The 90-day citation window.
Spreading thin content across too many pages
AI engines don't evaluate pages one at a time. They issue 8-12 fan-out sub-queries per search and look for domains with depth across those sub-topics. A site with 50 thin blog posts touching different keywords loses to a site with 15 deep, interlinked pages covering one subject from multiple angles. In CiteGap audits, the split-authority pattern (two or more pages on the same domain competing for the same AI query, neither building full citation authority) shows up in roughly half the brands we analyze.
Why depth beats volume in AI search: Topical authority and query fan-out.
Writing marketing copy instead of answers
AI engines retrieve 37 pages per query and cite about 5. The 32 that get cut almost always fail at the same point: the opening paragraph is brand messaging instead of a direct answer. "Our award-winning platform delivers best-in-class solutions" gets skipped. "The average mid-size CRM costs $45-85 per user per month, with implementation taking 6-12 weeks" gets cited. The ranking stage of the RAG pipeline rewards fact density and answer-first structure, not brand voice.
How the full retrieval-to-citation pipeline works: How ChatGPT decides what to cite.
Ignoring the mention-link gap
A brand sees their name in AI answers and assumes visibility is working. But appearing in the answer and earning the click are two different outcomes. Only 6-27% of mentioned brands also earn source citations with actual links. The rest get name-dropped while aggregators and competitors capture the click. This is the metric most brands don't know to track, and it is the one that explains why AI "visibility" isn't translating to traffic.
The full analysis of what causes this gap and why it matters: The mention-link gap.
Missing technical signals
Brands focus on content while the technical layer goes unchecked. Crawl access, structured data, and rendering method affect whether AI engines can even read your pages. Pages with the right technical signals are 30-40% more likely to be cited in AI answers. The irony: llms.txt, the signal everyone asks about first, has zero measurable correlation with AI citations according to a study of 300,000 domains. The signals that actually matter are less obvious.
Which technical signals actually move citation rates: The technical signals behind AI search.
Underestimating aggregator competition
Wikipedia accounts for 47.9% of ChatGPT's top citations. Reddit captures 46.5% on Perplexity. Five review platforms hold 88% of all review-platform links in Google AI Overviews. These sites don't have better expertise than brands. They have better content format for AI retrieval: comparison tables, structured Q&A, neutral tone, and third-party validation. Brands bring marketing copy to a data fight.
How aggregators outcompete brands and what the structural difference is: Why AI engines trust aggregators over brands.
The compounding problem
These eight mistakes don't operate independently. Stale content compounds with thin content. Marketing-first copy compounds with missing schema. Single-engine checking means you don't even see the problems on the engines where your competitors are winning.
That compounding is why self-diagnosis is unreliable. A brand can read all eight posts above, understand each concept, and still not know which of these problems is actually blocking their citations on which engine for which queries. The diagnosis requires per-engine, per-query, per-page data that connects the symptoms to root causes.
FAQ
How many of these mistakes does the average brand make? In CiteGap audits, the median is five out of eight. The most common combination is stale content, marketing-first copy, missing technical signals, and single-engine assumptions. Brands with strong SEO programs typically make fewer technical mistakes but still miss the content format and multi-engine gaps.
Can I check for these mistakes myself? You can identify some surface-level symptoms (Is my content older than 90 days? Do I have schema markup?). But the hard part is knowing which mistakes are actually blocking your citations on which engines for which queries. A page can have perfect schema and still not get cited because the content format fails the RAG pipeline. The interaction effects require diagnostic data.
Which mistake has the biggest impact on AI citations? It depends on the brand's starting position. For brands with no GEO strategy, the SEO-equals-AI assumption is the most expensive because it delays action entirely. For brands already investing in content, the mention-link gap and content freshness are usually the highest-impact findings.
Do these mistakes affect all AI engines equally? No. Stale content hits Perplexity hardest (50% of citations from the current year). Marketing-first copy affects ChatGPT most (only 15% of retrieved pages survive to citation). Missing technical signals affect Google AI Overviews disproportionately because of its reliance on schema and E-E-A-T signals. This engine-specific impact is why multi-engine audits matter.
How quickly can these be fixed? Technical signals (schema, robots.txt) can be implemented in days. Content freshness and format changes take weeks. Building topical authority takes months. The sequence matters as much as the speed, because fixing the wrong problem first wastes budget while the actual blocker remains.
CiteGap audits diagnose which of these eight patterns are blocking your brand, per engine, per query, with specific page-level fixes. Request a consultation.