A South African business that used to capture organic traffic for "best [category] in Johannesburg" queries has watched the SERP shift in 2025 and 2026. The answer is now an AI Overview, a ChatGPT citation or a Perplexity answer. The business either gets cited or it doesn't — and the click-through rate is lower whether it does or doesn't.
This post explains why AI search engines cite some businesses and not others. It covers the entity, schema, E-E-A-T and corroboration signals that drive AI citations, and the practical work a South African business can do to be cited.
The citation mechanic
AI search engines (Google AI Overviews, ChatGPT, Perplexity, Gemini, Bing Copilot) generate answers by combining three input types:
- The query. The user's question.
- The corpus. The web pages, knowledge bases and structured data the engine has indexed.
- The ranking signals. The entity recognition, authority, recency and corroboration signals the engine uses to decide which sources to cite.
The engine's job is to construct an answer that is correct, attributed and authoritative. The citation decision is the output of a ranking process applied to the corpus. A business that is not in the corpus cannot be cited. A business that is in the corpus but ranks poorly on the signals will not be cited.
The five signals that drive AI citations are below.
Signal 1: entity recognition
The engine must recognise the business as a coherent entity. This is the Entity Optimisation work. The signals:
- Schema.org structured data. Organization, LocalBusiness, Service and Product schema with consistent @id references. The schema 101 article covers the foundation.
- Knowledge panel presence. A Google Knowledge Panel for the business name signals entity recognition. Knowledge Panels are built from Wikipedia, Wikidata, schema, and authoritative directory listings.
- Wikidata entry. A Wikidata Q-number for the business is the strongest entity signal. Most SMEs do not qualify for a Wikidata entry, but a parent brand or founder with a notable profile may.
- Consistent NAP across the citation graph. Name, address, phone consistency across the web. The same NAP work that supports local SEO supports entity recognition.
A business with strong entity signals is in the engine's model. A business with weak entity signals is invisible to the engine as a coherent unit.
Signal 2: schema structured data on the cited content
When an AI engine cites a page, the citation is overwhelmingly drawn from the page's structured data, not from the visible prose. The schema types that drive AI citations:
- Organization. Sitewide. Defines the business entity.
- Service or Product. On the page being cited. Defines what the business sells.
- FAQPage. On the page being cited. Drives the answer-style content.
- HowTo. On instructional content. Drives step-by-step citations.
- Article. On blog posts. Drives authorship and date signals.
A page with rich schema is easier for the AI to extract a clean citation from. A page without schema forces the AI to parse prose, which is less reliable and less likely to be cited.
Signal 3: E-E-A-T signals
E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) is Google's framework for evaluating content quality. AI engines use similar signals. The components:
- Experience. First-person accounts, case studies, original research, photos and videos showing real-world work. The strongest E-E-A-T signal for service businesses.
- Expertise. Author credentials, professional qualifications, certifications, awards. The "About the author" page is the canonical place for this.
- Authoritativeness. Backlinks from authoritative sources, mentions in reputable publications, listings in industry directories.
- Trustworthiness. HTTPS, privacy policy, contact information, real address, real phone number. The basic trust signals.
A page with strong E-E-A-T signals is cited more often than a comparable page with weak signals. The lift is most visible on YMYL (Your Money or Your Life) topics — health, finance, legal — but applies across the board.
Signal 4: corroboration
The engine does not cite a single source. It cites sources that are corroborated by other sources. A business mentioned on its own website, on its LinkedIn, on its Google Business Profile, on its industry directory listings, and on a Wikipedia or Wikidata entry is more likely to be cited than a business mentioned only on its own website.
The corroboration work:
- Be listed on industry directories. Legal, medical, hospitality — every industry has its directories.
- Be mentioned in trade press. A mention in a reputable publication is a strong corroboration signal.
- Be reviewed on third-party platforms. Hello Peter, Google Reviews, Trustpilot — the platforms themselves are corroboration signals even before the review content is read.
- Be cited in Wikipedia or Wikidata. For notable businesses only. Most SMEs will not qualify.
A business with 5 to 10 corroborating sources is cited more often than a business with 1 (its own website). The compounding is significant.
Signal 5: recency and freshness
AI engines favour recent content for queries that imply recency. "Best [category] in 2026," "current pricing," "latest regulations." A page last updated in 2024 is deprioritised for these queries.
The freshness work:
- Quarterly content refresh. Update dates on existing content. Add 2026 to titles where applicable. Refresh statistics and examples.
- Active publishing. A site publishing weekly is treated as active. A site silent for 6 months is treated as stale.
- Schema dateModified. Article schema with current dateModified signals the engine that the content is fresh.
A business that publishes regularly and updates existing content has a measurable lift in AI citation rate for recency-implying queries.
The work to be cited
For a South African business that wants to be cited by AI engines:
- Audit the entity signals. Schema, knowledge panel, Wikidata, NAP. Identify the gaps.
- Ship the schema foundation. Organization, LocalBusiness, Service or Product, BreadcrumbList. The schema 101 article is the start.
- Build the E-E-A-T signals. Author bios, case studies, original research. The Topical Map work is the strategic frame for this.
- Earn the corroboration. Industry directories, trade press mentions, third-party reviews.
- Maintain the freshness. Quarterly content refresh, active publishing, schema dateModified.
A South African SME doing all five over 6 to 12 months sees a measurable lift in AI citations. The lift is most visible for non-branded queries where the engine has to choose which businesses to cite.
The connection to traditional SEO
AI citation work and traditional SEO work overlap heavily. The schema work, the E-E-A-T work, the corroboration work and the freshness work all benefit traditional SERPs as well as AI citations. The investment pays back in both channels.
The difference is in measurement. Traditional SEO is measured by ranking, traffic and CTR. AI citations are measured by being cited — the engine has to surface the business name, not just rank the page. A new measurement framework is needed: prompt testing, citation tracking, AI Overview presence tracking.
The work is the same. The measurement is different.
FAQ
How do I know if my business is being cited by AI engines?
Prompt the major AI engines with "[your category] + [your city]" and similar queries. Note which businesses are cited. A more rigorous approach is to use an AI visibility tool that tracks citations over time.
Will AI search kill traditional SEO?
No, but it will reduce click-through. A business ranked position 1 in a SERP with an AI Overview still gets traffic, but less than it would have without the AI Overview. The work to be cited inside the AI Overview compensates for the lost click share.
How long until the AI citation work shows results?
3 to 6 months for the entity and schema work. 6 to 12 months for the E-E-A-T and corroboration work. AI citation work compounds similarly to traditional SEO — slower start, accelerating return.
Does my industry matter?
Yes. YMYL industries (health, finance, legal) have higher E-E-A-T requirements and lower AI citation rates without strong signals. Lower-stakes industries (eCommerce, hospitality) have higher AI citation rates with less work.
Should I block AI crawlers?
No. Blocking GPTBot, ClaudeBot or Perplexity-Bot removes the business from the corpus and guarantees zero citations. Allow them in robots.txt. The citations are net positive for brand visibility even if the click-through is lower.
If you want a second opinion on the AI citation work for a specific business, send us the URL and the queries you want to be cited for. Most entity optimisation engagements are scoped as 3 to 6 month projects with quarterly milestones, focused on the schema foundation first and the E-E-A-T work layered on top.