[[FIG:kb-0406]]

The crisis sequence, adapted for AI

1. Trace the source. Identify where the article originated and how it is being picked up. AI-fabricated content that ranks in search can itself become a source the engines go on to cite.
2. Prepare authoritative counter-content. Produce clear, well-sourced material that addresses the claims directly and gives the engines accurate alternatives to weigh. Engines resolve conflicting information by weighting sources on authority and recency, so the counter-content has to be credible enough to outweigh the viral piece, not just exist alongside it.
3. Run platform-policy escalation in parallel. If the article sits on a platform with relevant policies, run the policy-based escalation paths alongside the content response, not after it.
4. Monitor amplification across the engines. Track how the eight AI engines AIQ currently tracks are absorbing the viral content: which engines are picking it up, which sources they are pairing it with, and how the narrative moves over time.

The AI-specific dimensions

• Engines synthesize; they do not republish. Each engine builds its answer from the surrounding source ecosystem, weighted by authority and recency. That is why you target the sources, not the engine outputs.
• A single viral source can be over-weighted. Engines can lean too heavily on one contested source when forming an answer, so well-sourced counter-content gives them a higher-authority alternative to weigh.
• Amplification crosses platforms. Viral content can spread beyond where it started, including into branded search, so monitor across the engines AIQ tracks rather than the originating platform alone.
• The clock is hours and days. Retrieval-driven engines reflect newly published sources within minutes to hours, so both the counter-content and the monitoring have to keep pace.

The eight engines AIQ currently tracks are ChatGPT, Gemini, Copilot, Perplexity, Claude, Grok, Google AI Overviews, and Google AI Mode.

The post What happens when an AI-generated article about your company goes viral? appeared first on Five Blocks.

[[FIG:kb-0362]]

Step 1, Pillar content: cover core topics in depth

Pillar pages cover the brand’s core topics thoroughly, with named expert authorship and clean structure. The HubSpot topic-cluster model describes the linking dynamic: supporting pages link back to the pillar, and that linking action signals to search engines that the pillar page is an authority on the topic, building ranking and citation potential over time. The same signal applies to AI engines, which read pillar-and-cluster architecture as evidence of topical depth rather than scattered coverage.

Step 2, Supporting content: write for extraction

Supporting pages answer the specific questions readers ask about each pillar, written for the extract:

• Question-format headings: H2 and H3 headings phrased as the actual question, in natural language rather than marketing copy.
• Direct answer immediately below each heading, self-contained enough to be lifted without context.
• FAQPage schema where the format fits, so the structure is machine-readable. The GEO framework (Aggarwal et al., KDD 2024) frames this as optimizing web content for visibility in generative engine responses: engines synthesize answers from retrieved sources, and content built to be quoted is more likely to be retrieved and cited.

Step 3, Internal linking: tie the cluster together

Each supporting page links back to its pillar, and the pillar links forward to the supporting pages. This architecture lets the engines read the relationships clearly, treating the cluster as a coherent body of work on a topic rather than a set of independent articles. Pages not connected to the cluster receive weaker topical-authority signal regardless of their quality.

Step 4, Freshness: update on a regular cadence

A consistent update cadence keeps the freshness signal positive rather than letting it decay: revise pillar pages as topics change, add supporting pieces as new questions come up, keep publication dates current. This is maintenance work, not a one-time build, and it reinforces the architectural work above.

What sustained execution produces

Over six to twelve months, this combination makes the engines recognize the domain as authoritative on the relevant topics, which leads to citation in AI responses. Scattered content published without structure performs poorly regardless of volume.

The post How do you build a content strategy specifically for AI visibility? appeared first on Five Blocks.

[[FIG:kb-0350]]

Correction workflow by source type

1. Identify the anchor source.
   Run the relevant prompts through AIQ across the eight engines AIQ currently tracks: ChatGPT, Copilot, Gemini, AI Overview, Perplexity, Grok, Claude, and Google AI Mode. For retrieval-heavy engines (Perplexity, ChatGPT Search, Gemini with Google Search grounding), the cited sources appear inline. For training-baselined engines, compare the exact phrasing of the misinformation against Wikipedia, aggregators, and major press to find the likely source.
2. If Wikipedia is the anchor: file a Talk-page edit request.
   Wikipedia is one of the most heavily weighted sources across AI engines and a primary input to Google’s Knowledge Graph. Corrections go through the standard disclosed-COI edit-request process on the article’s Talk page: quote the wrong text, propose the exact replacement, and cite reliable secondary sources. Direct edits from a conflict-of-interest account get reverted even when the correction is accurate. Done correctly, the Talk-page process produces durable corrections that the engines then absorb.
3. If a specific article or aggregator is the anchor: pursue a press correction or strengthen competing sources.
   Most reputable news outlets have published correction processes, and documented factual errors submitted through the right channel are corrected more often than practitioners expect. Where a correction is not available, build authoritative competing sources that describe the same facts accurately: press placements in outlets the engines weight, and updated owned content, until the engines re-weight away from the inaccurate anchor.
4. If a Knowledge Graph or Wikidata value is wrong: submit corrections through the appropriate channel.
   Google’s Knowledge Graph can be corrected through verified entity feedback, available once identity is verified via the Knowledge Panel claim process, and indirectly through Wikidata corrections, which flow into the Knowledge Graph. Wikidata updates reach AI engines faster than Wikipedia narrative changes and should be a priority for structured-data errors such as wrong founding dates, leadership, or corporate relationships.
5. Monitor propagation across the eight engines AIQ tracks.
   Once source-level corrections are in place, track daily in AIQ until each engine reflects the accurate information. Retrieval-heavy engines, which perform live web searches when a query arrives, typically update within days of the underlying source changing. Training-baselined engines update only on retraining cycles, which can take longer. The engines do not all move at once, so monitoring across the eight AIQ tracks stops you from trusting one engine’s correction while others still carry the error.

What cannot be done

No service or technical capability lets anyone, including the companies that build the engines, edit a specific AI response directly. The work is always at the source layer. Any firm claiming direct control over AI output is misrepresenting what the discipline can do. Source-level corrections, sustained over time and monitored through AIQ, are what produce durable results.

The post How do you correct AI-generated misinformation about your brand? appeared first on Five Blocks.

[[FIG:kb-0348]]

The core components

Wikipedia article

The keystone for any subject that meets the general notability standard (significant coverage in multiple reliable, independent, secondary sources). Wikipedia is a primary data source for the Google Knowledge Graph, a frequent retrieval target in AI-generated answers, and the source AI engines are most likely to paraphrase closely when summarizing a subject.

Wikidata entry

The machine-readable twin of the Wikipedia article. AI engines and Google’s entity layer query Wikidata directly for structured facts, founding dates, leadership, headquarters, parent and subsidiary relationships, regulatory identifiers. Wikipedia language versions across roughly 300 languages are linked through Wikidata to the same underlying entity record.

Schema markup on owned properties

Organization or Person schema on the brand’s own pages, with sameAs properties linking to the canonical identifiers (Wikipedia URL, Wikidata Q-ID, official social profiles). These links tell the engines that the page belongs to the same entity already in the Knowledge Graph, and they help disambiguate common or shared names. AI engines and Google’s entity layer read schema markup directly as a signal about what a page is and how it relates to entity context.

Authoritative third-party citations

Mainstream press, industry registries, regulatory filings, and other independent sources that confirm the same entity facts. They reinforce the entity signal, and the engines weigh them alongside the structured knowledge layer.

Attribute consistency

The same name, affiliation, founding date, and relationship descriptions everywhere, across Wikipedia, Wikidata, schema markup, owned content, and third-party sources. Inconsistency is one of the main reasons AI engines return contradictory answers about the same entity across different prompts or engines.

Why it matters for AI responses

Google’s AI Overviews and Gemini weigh Wikipedia and the Knowledge Graph as primary sources when summarizing a subject. AI engines disambiguate common names using entity infrastructure: Wikipedia disambiguation pages, unique Wikidata IDs, and schema sameAs links. Companies with multiple operating brands are the most exposed to fragmentation, the engines give parent and subsidiary brands unrelated descriptions when the entity infrastructure does not spell out the relationship. A complete, consistently maintained entity layer is the most durable lever for shaping what AI systems say about a brand.

The post What is entity optimization for AI? appeared first on Five Blocks.

[[FIG:kb-0305]]

The four disambiguation layers

1. Wikipedia disambiguation pages

   When several distinct subjects share the same name, Wikipedia creates a disambiguation page that lists each subject and links to its article. This is one of the primary signals engines use to tell apart entities with identical or near-identical names. Without a disambiguation page or a clearly differentiated article title, the entity is harder for an engine to separate from its namesake. Wikipedia’s own documentation defines a disambiguation page as a non-article page that lists the various meanings attached to a name and links to the articles that cover each one.

2. Wikidata unique identifiers (Q-IDs)

   Every Wikidata item carries a unique Q-ID, a number prefixed with the letter Q, such as Q42 for Douglas Adams, that anchors the entity regardless of how many others share a similar name. Wikidata’s own introduction defines items as uniquely identified by Q-numbers. Each language-version Wikipedia article for that entity is linked as a sitelink on the same Wikidata item, which gives AI engines and the Google Knowledge Graph a machine-readable identity anchor to query. A missing or incomplete Wikidata entry removes that anchor and leaves the engine relying on weaker, more ambiguous signals.

3. Schema.org Person markup with sameAs links

   Schema.org’s sameAs property lets a brand’s owned web pages declare a reference URL that “unambiguously indicates the item’s identity”, usually the entity’s Wikipedia article, Wikidata entry, or official website. When this markup is present on a Person or Organization page, it gives AI engines and the Knowledge Graph an explicit machine-readable signal connecting the page to the canonical entity. This is the layer where owned infrastructure asserts identity directly rather than waiting for engines to infer it from surrounding text.

4. Contextual cues in the query

   Industry descriptors, geographic qualifiers, role titles, and co-mentioned related entities in the user’s query all act as soft disambiguation signals. These cues help when the structural infrastructure above is strong, but they cannot fully make up for a missing Wikidata entry or absent schema markup when the name overlap is close. Contextual cues are what the engine falls back on when the machine-readable layers do not resolve the ambiguity cleanly.

The conflation failure mode

When entity infrastructure is weak, meaning no Wikidata entry, no schema markup, and no clean Wikipedia disambiguation page, AI engines are more likely to conflate the target with another entity that shares the same or a similar name. Research on large language models shows that models can consistently mishandle broad classes of human names when the disambiguating contextual cues are absent. The fix is upstream: a complete Wikidata item with sourced statements and sitelinks, schema markup with sameAs pointing to Wikipedia and Wikidata, and a clear Wikipedia disambiguation page or distinct article title. These are infrastructure interventions. Prompt-layer workarounds do not address the root cause.

The post How do AI models handle disambiguation for people and companies with common names? appeared first on Five Blocks.

[[FIG:kb-0356]]

Which platforms matter and for what

• Reddit: the most-cited social domain across AI engines. Threads surface for evaluative, comparative, and experiential queries: “is X a good service,” “what is it like to work at Y,” “how does Z compare to competitors.” AI engines weight these community discussions for opinion and reputation queries.
• YouTube: AI engines crawl and embed video transcripts, treating them like written articles. YouTube is a clear favorite in Perplexity (16.1% of citations) and Google AI Overviews (9.5%), particularly for tutorial, product comparison, and explainer queries.
• LinkedIn: one of the top-cited domains in AI-generated answers per the Peec AI study. The platform’s high domain authority makes it a consistent source for professional and company-related queries. Note: the claim that individual LinkedIn posts by executives are routinely cited for executive perspectives lacks a verified source in our library and should be treated as plausible but unconfirmed.
• X (formerly Twitter): X threads appear as sources in some retrieval-based AI engines, particularly for recent events and opinion. How much X gets cited varies by engine and query type, and current studies offer limited direct evidence of consistent citation frequency.

What this means for reputation work

A brand’s social presence, and its key people’s social presence, is now reputation infrastructure rather than only an engagement channel. A platform such as Glassdoor or Reddit where a brand’s reputation is discussed is one of the channels AI engines draw from when they answer questions about that brand. When a brand is absent or poorly represented in those discussions, other voices fill the space. Monitoring and managing these layers belongs inside an AI reputation program.

The post What is the relationship between social media presence and AI search results? appeared first on Five Blocks.

Visibility vs. reputation in AI responses

[[FIG:kb-0329]]

Why visibility alone is not enough

Major AI engines draw on different source sets. Ahrefs found that 86% of the top sources cited by ChatGPT, Perplexity, and AI Overviews are not shared across all three, so a brand’s AI presence is never uniform. One engine may cite the brand from a favorable press piece; another may pull from a less flattering secondary source. High visibility across all three engines does not guarantee that the narrative is consistent or accurate in any of them.

The GEO paper (Princeton et al., ACM SIGKDD 2024) frames the optimization target as increasing impression in generative engine responses. Impression is necessary but not sufficient. The comms objective is a correct, favorable impression, which means managing the source layer, Wikipedia, owned properties, authoritative press, structured data, not just appearance frequency.

The practical distinction

Visibility is the marketing read. It answers: are we present in AI answers for our category? Reputation is the comms read. It answers: when we are present, what is the AI saying, and is that what we want said? A full AI reputation program has to work both questions. A brand that optimizes only for visibility while ignoring what the engines say has done the easier half and left the rest undone.

The post How is reputation different from visibility in GEO? appeared first on Five Blocks.

[[FIG:kb-0351]]

The six attributes of citable content

Fact density

Named entities, concrete numbers, real dates, and specific claims. Vague or hedged prose gives the engine nothing to quote. Retrieval-augmented engines in particular pull self-contained, directly answerable passages.

Clear structure

Headings that frame each question, a direct answer immediately beneath, and lists or tables for enumerable content. Google’s AI optimization guidance says pages organised by paragraphs, sections, and headings give generative features a clearer structure to work with. Schema markup (Article, Person, Organization, FAQ) makes that structure machine-readable and ties the page to the correct entity via sameAs links to Wikipedia and Wikidata.

Authoritative sourcing

Every non-trivial claim carries a citation to a source the engines themselves treat as credible. The Princeton GEO study (ACM SIGKDD 2024) found that adding citations, quotations, and statistics to content improved visibility in AI engine responses by 30-40% compared to baseline. Engines weight sources by domain reputation and citation patterns, so third-party corroboration matters more than adding more owned pages.

Recency

Real publication and last-updated dates, and references that have not gone stale. Retrieval-first engines such as Perplexity rank pages partly by recency alongside domain authority and topical relevance, so a recently updated authoritative article often outranks an older one on the same topic.

Domain authority

The hosting domain’s established credibility. AI engines weight sources partly by domain reputation and how often a domain appears in authoritative citation patterns. Content on a high-authority domain has an advantage regardless of individual-page quality.

Explicit authorship

A named expert with bio context that makes the expertise verifiable. Google’s helpful-content guidance ties E-E-A-T signals (experience, expertise, authoritativeness, trustworthiness) directly to how content is evaluated. Clear, credentialed authorship is one of those signals.

What happens when one dimension is missing

Content that fails on any of these can still be useful to human readers but is less likely to shape AI synthesis. A well-structured, well-sourced page on a low-authority domain will lose to a comparable page on a high-authority one. A fact-dense page with no clear structure or schema is harder for engines to parse and attach to an entity. The traits reinforce each other, and the highest citation probability comes from content that meets all six.

The post How do you create content that AI models prefer to cite? appeared first on Five Blocks.

[[FIG:kb-0399]]

Step 1: diagnose the source the engine is leaning on

Retrieval-based engines expose the sources behind an answer, so the first move is to read which source is carrying the competitor’s recommendation. Each source type points to a different fix.

Step 2: do the concrete work per source type

• Directories and ranking guides: strengthen the brand’s presence in the ones the engines weight, so the brand appears where ranked recommendations are drawn from.
• Comparison content: produce authoritative comparison material where the existing content is dated or one-sided, giving the engines new, better-sourced reading.
• Entity infrastructure: build the Wikidata entry, schema markup, and Knowledge Panel signals that let the engines resolve the brand as a peer when they decide which firms to name.
• PR coordination: work with PR on placements in earned, third-party coverage the engines weight, not wire releases alone.

Why diagnosis comes first

The approach works when the source diagnosis is correct, because each source type calls for different work, and engines can keep serving an outdated comparison or over-weight a single source long after it should have moved on. What does not work is arguing with the engines about their recommendations: you can’t edit model outputs directly, so the influence comes from shaping the sources they draw on.

The post How do you manage reputation when AI tools recommend competitors over you? appeared first on Five Blocks.

[[FIG:kb-0357]]

Step 1: Mark up Q&A content with FAQPage schema

FAQPage schema marks question-answer pairs as machine-readable, so AI models and assistant platforms can identify the content as structured Q&A and extract it cleanly. Google Search Central confirms that properly marked-up FAQ pages “may be eligible to have…an Action on the Google Assistant,” which ties the markup directly to assistant selection. The Schema.org FAQPage type defines the page as one presenting one or more “Frequently asked questions”, a signal the engine can act on without further interpretation.

Note: Google retired FAQ rich results in 2026, so FAQPage markup no longer produces a visual accordion in the standard Google SERP. The markup still helps because it makes page structure machine-readable for AI models (including Google’s own AI Overviews) and for other assistant platforms that did not adopt the same policy.

Step 2: Front-load concise definitional answers

Voice assistants and AI engines resolve entity and definitional queries by retrieving the clearest, most authoritative passage that answers the question. A clean two-to-three-sentence answer at the top of a topic page, before the surrounding context, cuts the interpretive work the engine has to do. Google’s featured-snippet and AI Overview systems use closely related selection logic and favour clean, extractable answers backed by source authority; pages that meet this bar tend to get cited more often across AI answer engines. The claim that definitional content “wins” these slots outright is not independently sourced. The mechanism (extractability plus source authority) is supported, but placement alone is not a guarantee.

Step 3: Use HowTo schema for procedural queries

Procedural voice queries (“how do I…”, “what are the steps to…”) are a distinct pattern. Numbered steps plus HowTo schema let AI engines extract the procedure cleanly: Schema.org defines HowTo as “instructions that explain how to achieve a result by performing a sequence of steps,” and Google Search Central confirms that structured data is “a standardized format for providing information about a page and classifying the page content.” Whether voice assistant platforms select HowTo schema at higher rates than plain prose is not independently supported in the research. The supportable basis for the recommendation is the extractability the markup provides.

Step 4: Build strong entity signals (Wikidata and Knowledge Panel)

Before an AI assistant can pick a piece of content as the right answer, it has to identify the entity the query is about. That identification relies on machine-readable entity signals. Wikidata is a free, collaboratively edited multilingual knowledge graph that “allows digital assistants (Siri, Alexa, Google) and bots to add and pull information from Wikidata and present it to their users.” A Google Knowledge Panel is “a primary source for direct answers” in voice search, not merely a search result. An accurate, complete Wikidata entry and a Knowledge Panel that reflects current facts are the base layer that makes the content-level steps above work.

Step 5: Connect the pieces with schema sameAs links

Schema markup on the site itself (Organization, Person, Article types) can carry sameAs links that point to the entity’s Wikidata entry, Wikipedia article, and official profile pages. This ties the entity together across the web and connects it to the canonical identifiers AI engines use for disambiguation. The Schema.org sameAs property is defined as a “URL of a reference Web page that unambiguously indicates the item’s identity,” citing Wikipedia and Wikidata as explicit examples. With accurate Wikidata and Knowledge Panel data, this gives the assistant a coherent, cross-referenced picture of the entity when it resolves a query.

The post How do you prepare for voice search and AI assistants? appeared first on Five Blocks.