Where Do AI Engines Get Their Information and How Can Your Brand Be Included?

What changed in how AI engines find information?

AI engines have made citation visibility a broader web problem. A brand can publish clear, useful pages and still be absent from an answer if the wider set of sources an engine draws on does not establish its relevance, authority or category fit. The shift is from asking whether a page can rank to asking whether a brand has enough trustworthy evidence to be cited when an answer is assembled.

The Q3 2026 Mencoro study examined 23,466 AI answers across ChatGPT, Google AI Overviews, Google AI Mode and Perplexity. It recorded 176,332 citations across 22,193 domains for 42 tracked brands in eight sectors in Spain. That is not a census of every answer on the web. It is a substantial, defined observation of commercial buyer questions, and it gives marketers a useful warning: no single source class owns AI visibility.

The old instinct was to treat the company website as the centre of every visibility strategy. The data does not support treating owned media as sufficient. Mencoro reports that a tracked brand's own website received 4.8% of all citations in its sample. Your site remains where you control product facts, service detail and evidence. It is not the whole citation surface.

Google describes AI Overviews as AI-generated snapshots with links for people to explore further. It also says AI Overviews may appear when its systems determine generative AI can help people understand information from a range of sources. That does not provide a formula for being cited. It does confirm the important operating reality: a brand should make its facts easy to verify wherever a prospective buyer may encounter them.

1. Before this evidence, teams often optimised the brand site and assumed authority would flow outward.
2. After this evidence, teams should maintain the brand site as a factual source, then build a credible footprint across the category surfaces that AI answers actually cite.
3. What to do: measure citations by engine and prompt set, identify recurring cited source types, then close evidence gaps without trying to manipulate model outputs.

Where did the Q3 2026 study find AI citations?

AI citations in the study clustered around commercial and category-specific information, not just national news coverage. Mencoro reports that 72.6% of citations went to company websites, including competitors, shops, clinics and manufacturers. Directories and marketplaces received 11.6%, while news and magazines received 4.3%. For commercial questions, functional market information had far more citation volume than general press.

That finding changes the first research task for a growth team. Do not start with a generic list of prestigious publications. Start with the pages and platforms a buyer uses to compare options, verify availability, understand a service or check a provider. In one vertical, that may mean a respected trade publication. In another, it may mean a marketplace, specialist directory, manufacturer page or public record.

The source mix does not make earned media irrelevant. Press can establish a public record and can be particularly relevant where a question needs context. The lesson is narrower and more useful: publication prestige alone is not a reliable substitute for category coverage. A citation strategy should follow the question, the engine and the specific evidence a buyer needs.

Omnicite calls the result Citation Share, the percentage of relevant AI answers in a category that cite your brand. Citation Share gives a team a way to observe the outcome rather than assume that traffic, rankings or domain authority automatically translates into AI visibility. Rankings got you found. Citations get you chosen.

1. Company websites received 72.6% of citations in the Mencoro sample.
2. Directories and marketplaces received 11.6% of citations in the same sample.
3. News and magazines received 4.3% of citations in the same sample.
4. The tracked brand's own website received 4.8% of all citations in the sample.

Is there one list of websites every brand needs?

No. The study found that 57.4% of citations pointed to domains cited for only one brand. That pattern argues against a universal placement checklist. A source that is highly relevant for a legal service, retailer or travel business may have no role in a technology buyer question. The useful unit is the local citation ecosystem around a category, not a generic authority score.

This is good news for brands that lack the budget or need for broad publicity. A specialist source can matter because it answers a specialist question. The work is more exacting than buying a list of placements. It requires mapping which prompts matter, which sources appear, what facts those sources contain and whether your own information is consistent enough to withstand comparison.

Every citation deserves close reading as evidence of an information route, not as a trophy. When an answer repeatedly cites a category directory, examine the information it exposes. Where an answer cites an official record, confirm that your relevant facts are accurate and current. For comparison resources, identify whether your brand is absent because of a coverage gap, a factual gap or an editorial decision you should respect.

A broad strategy can still have a disciplined shape. Build authoritative pages on your site. Maintain factual profiles where buyers expect to find them. Contribute evidence to relevant third-party sources when there is a legitimate route. Refresh details when your has changes. This is Citation Engineering, not a promise that any engine will cite a page on demand.

1. Map the buyer questions that matter to your category.
2. Record which domains appear for each question and engine.
3. Classify each source by its role, such as company site, directory, marketplace, public body or editorial publication.
4. Fix factual coverage before pursuing additional third-party visibility.

Do ChatGPT, Google AI and Perplexity cite the same sources?

No. Mencoro found different source profiles across engines, which means a single content format is unlikely to cover every citation opportunity. Google AI Overviews and AI Mode showed a much stronger presence for social and visual platforms in the Spanish sample. ChatGPT had a stronger association with written press and official records. Perplexity stood out for YouTube citations.

The social signal was especially clear in the study. Instagram was cited for 40 of 42 brands, and Mencoro reports that 97.6% of those Instagram citations came from AI Overviews and AI Mode. TikTok had no ChatGPT citations in the data set. These are observations from a defined place and period, not a platform rule. They should guide testing, not justify copying a tactic across every market.

The text and records pattern also needs a careful reading. Mencoro reports that ChatGPT accounted for 84.1% of citations to El País and 94.4% of citations to Spain's official gazette, the BOE. It reports that public bodies made up 11.7% of ChatGPT citations. The lesson for brands is not to manufacture official-looking language. It is to make claims traceable to proper documentation and to avoid leaving important public facts vague.

Perplexity accounted for 191 of 246 YouTube citations in the sample, or 78%. A brand with a useful video library should see that as a reason to ensure each video is accurate, specific and connected to real buyer questions. Video alone does not establish expertise. Clear titles, visible product context and supporting written material make the information easier for people to assess as well.

1. For Google AI Overviews and AI Mode, review whether accurate visual and social information exists where the category expects it.
2. For ChatGPT, strengthen primary documentation, clear written explanations and public evidence.
3. For Perplexity, evaluate whether video answers real questions and aligns with the brand's written facts.
4. For every engine, track observed citations instead of assuming one channel produces the same result everywhere.

Who does this change affect most?

This change affects B2B SaaS and technology growth teams that do not know whether ChatGPT recommends them or a competitor for a category question. It also affects local, multi-location and service businesses whose buyers now ask AI for the best provider in a city. In both cases, the risk is not merely lower traffic. It is exclusion from the short answer that shapes a buyer's first impression.

A SaaS team may have strong product pages yet lack comparisons, implementation evidence or third-party category coverage. A local service business may have an accurate site while its listings, operating details or local references are incomplete. Neither problem is solved by publishing more generic blog posts. The missing evidence must be identified before content volume becomes useful.

The study also matters to brand teams that report SEO as a single aggregate number. AI visibility has several distinct dimensions. Citation Count per day measures volume. Answer Presence measures breadth across a question universe. Share of Voice measures the relative position against competitors. Citation Share is the headline measure of how often relevant answers cite the brand. Each answers a different management question.

Do not confuse a citation with a conversion or a guarantee of a sale. A citation is an observed inclusion in an answer. It can tell you where a brand appears and which sources are being surfaced. Teams still need to connect that visibility to their own enquiry, signup, booking or revenue measurement before making commercial claims.

1. B2B SaaS teams should inspect category, alternative and comparison prompts.
2. Local businesses should inspect service-plus-location prompts and AI Overviews presence.
3. Service businesses should verify listings, category descriptions and local facts.
4. Editorial teams should prioritise citable evidence over undifferentiated publishing volume.

How should a brand respond without trying to game AI engines?

Respond by improving quality, coverage and freshness across the evidence a buyer can check. Do not try to hack, game or manipulate models. Engines change, prompts vary and citation patterns are not a stable loophole. The durable work is simpler: publish authoritative content, maintain accurate facts and earn legitimate coverage in the places that make sense for the category.

Start with a prompt inventory. Write down the questions a buyer asks before and after they know a brand name. Include category questions, comparison questions, use-case questions and location questions where relevant. Run those questions across the engines you care about. Save the cited domains, the cited pages, the answer framing and the competitors that appear. This turns a vague AI visibility concern into a research backlog.

Next, compare the sources with your factual footprint. Your primary site should explain what you do, who it is for, what it does not do and how a buyer can verify important claims. Third-party information should be consistent with those facts. Where a specialist directory or marketplace is central to your category, confirm your presence is complete and accurate. Where independent editorial coverage is appropriate, bring useful evidence rather than a demand for a mention.

Finally, establish a measurement cadence. Citation visibility is an outcome that can shift as engines, sources and pages change. Track the same prompt set over time, split results by engine and review material changes. That is how a team can improve its Citation Share without pretending it controls an answer engine.

1. Define the category and comparison prompts that reflect real buying intent.
2. Capture citations and cited domains for each engine.
3. Repair inaccurate, incomplete or stale primary and third-party facts.
4. Publish source-backed material that answers a narrow buyer question clearly.
5. Measure Citation Share, Answer Presence and Share of Voice over time.

What should brands do first after this change?

First, audit the evidence behind your most important answers. The immediate goal is not to produce more content. It is to identify whether a buyer asking a relevant AI question can find clear, current and independently supported information about your brand. A precise audit often reveals that the gap is a missing product explanation, incomplete profile, outdated detail or absent category proof.

Second, separate owned content from the broader citation system. Your website should remain the clearest account of your offer. The Mencoro findings show why it cannot be the only account. A credible external footprint gives engines and buyers more routes to understand the category, compare options and verify a claim. That footprint must be earned and accurate, not mass-produced.

Third, make the before-and-after decision explicit. Before this evidence, a team might have measured rankings and assumed visibility would follow. After it, the team should measure citations in the answers buyers actually receive, then investigate the sources behind those answers. That is a more direct operating model for a world where there is no page two in an AI answer.

The aim is not omnipresence. It is useful inclusion on the prompts that matter to your market. Brands that build verifiable information across their primary site and relevant external sources give themselves a stronger basis for citation. They also give buyers a better reason to trust what an engine shows them.

1. Before this evidence, report rankings and traffic as the main proof of search visibility.
2. After this evidence, add citation measurement across relevant ChatGPT, Perplexity, Gemini and Google AI Overview prompts.
3. What to do now: audit source coverage, correct facts and track the resulting Citation Share.

Sources

Source: Mencoro

Mencoro analysed 23,466 AI answers and recorded 176,332 citations across 22,193 domains in Spain during Q3 2026. Mencoro, 2026-09-30

Source: Mencoro

The cited-domain study reports source distribution, engine-level differences and the tracked brand's 4.8% share of citations from its own website. Mencoro, 2026-09-30

Source: The AI Journal

The AI Journal reported the Mencoro Q3 2026 findings, including the study scope and its implications for commercial AI search visibility. The AI Journal, 2026-10-05