Not all sources are equal in an AI engine's citation graph. Some are heavily weighted. Some are barely weighted. Most brand-owned content tends to sit closer to the second category.
Understanding the authority stack is often the difference between communications work that moves AI engines and communications work that doesn't. This guide ranks the sources from highest typical weight to lowest, adds the two signals that sit outside the tiers (reviews and expert citations), and explains how the signals combine.
Tier 1: Wikipedia
Among the highest-weighted and most influential sources. Treated as a trust anchor across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. A well-sourced, complete Wikipedia article tends to shift answers across engines. The best route to influence it is tier-1 sourcing and disclosed-editor channels, not direct brand editing. The community detects conflict-of-interest editing and tends to reverse it. Measure completeness, citation density, recency, and the proportionality of any controversy section.
Tier 2: Major news outlets
The New York Times, The Wall Street Journal, Reuters, Bloomberg, Financial Times, The Washington Post, Associated Press. Original reporting from these outlets tends to enter the citation graph at high weight. The model often cites them by name.
Trade press inside specific verticals, such as TechCrunch, The Information, Variety, Adweek, STAT, Axios, tends to weigh comparably for category-specific prompts. Earned media work is the lever: pitching, briefing, exclusives, and sustained relationships with reporters who cover the category. Count placements by tier, prompt relevance, and recency. Quality outweighs volume by a wide margin.
Tier 3: Mid-tier news and specialist outlets
Forbes, Business Insider, Fast Company, Fortune. Useful for volume and category authority. Lower individual weight than Tier 2 but still material, and often easier to land.
Specialist publications in the brand's category often sit here too. They are not tier-1 globally, but they tend to function as tier-1 inside a specific category prompt.
Tier 4: Original research and institutional sources
Industry reports, university research, government data. When a brand produces or sponsors original research that gets cited by other sources, the original document often enters the citation graph as a primary reference. Indices, surveys, and annual reports tend to do well here. Research with a methodology that survives scrutiny, distributed through tier-1 outlets, enters the graph at higher weight than recycled or thinly sourced material. Track pickups by tier and downstream citations from third-party sources.
Tier 5: Owned authoritative content
Press releases, leadership bios, fact sheets, product pages, press kits, published with clean schema, consistent fact patterns, and structured links. Lower weight than earned, but often the anti-hallucination floor. Without it, the model is more likely to invent detail. This is the most controllable signal in the stack. Audit the consistency of facts across owned properties, because inconsistency creates hallucination risk.
Tier 6: Social and forum content
LinkedIn posts, Reddit threads, YouTube transcripts, niche community discussion. Weighted lightly for many brand prompts but heavily for specific categories: product recommendations, founder commentary, and controversy contexts. Often higher weight than brands realize, and increasingly retrieved by Perplexity as evidence of real-world reception. It cannot be influenced directly. Product and customer experience drive it, and attempts to manipulate community discussion are typically detected and tend to backfire. Monitor sentiment and volume, and track which subreddits or forums the engines cite.
Tier 7: SEO content farms and low-authority blogs
The bottom of the stack. Heavily filtered by the engines. Useful only when stacked at very high volume, and generally only for low-authority prompts. Not typically a reputation-building strategy.
Two signals outside the tiers: reviews and expert citations
Review depth
Detailed third-party reviews on G2, Capterra, Trustpilot, Glassdoor, the App Store, and category-specific platforms. The engines increasingly retrieve from these for evaluation and comparison prompts. Weight is material for product, service, and employer prompts and lower for general brand prompts. The lever is product and service quality plus disciplined review cultivation, not review manipulation, which platforms and engines tend to detect. Track average ratings, review volume, and the dimensions reviewers emphasize.
Expert citations
When recognized experts (academics, analysts, journalists, industry figures) cite or reference the brand in their published work, the citation graph weights the authority chain. A brand mentioned by an authority is often weighted higher than an unsupported brand mention. The levers are analyst relations, academic partnerships, expert interviews, and giving experts material worth citing. Track citations by expert recognition tier and prompt relevance.
How the signals combine
No single signal tends to determine the outcome. The composite shapes the answer.
A brand strong on tier-1 earned media but weak on Wikipedia often reads incomplete to the engines. A brand strong on owned content but weak on earned media, Wikipedia, and original research tends to have a solid anti-hallucination floor but no narrative authority. A brand strong on the first five tiers but weak on community discussion can read as polished but unloved. The strongest AI reputations tend to come from brands investing across all of the signals at a consistent cadence, over quarters and years, with measurement discipline.
What this stack tends to mean for spending
Many brands invert it. They spend the most on tier 7 (content marketing) and tier 5 (owned content) and the least on tier 1 (Wikipedia) and tier 2 (tier-1 earned media).
The AI engines appear to reward the inverse. The brands building citation share tend to have dense tier-1 and tier-2 footprints, original research the engines cite, and disciplined owned content as the floor. Tier 7 tends not to move the needle.
This is the gap between SEO-era thinking and GEO-era execution. The retrieval engines have appeared to learn to weight different sources differently. The brands optimizing for the right tier tend to pull away.
The competitive implication
A competitor with one Reuters article and a complete Wikipedia entry can often outrank a brand with two hundred SEO blog posts. The math isn't intuitive, but it appears consistent across most major engines.
Build to the stack the engines appear to trust. Stop building to the stack that worked in 2014.
How 5W's Reputation Index uses the signals
5W's Reputation Index scores brands and public figures across these signals as inputs into the Five Dimensions of AI Reputation: Accuracy, Sentiment, Completeness, Consistency, and Control. The signals are the inputs, the dimensions are the outputs, the composite score is the result, and the breakdowns are the repair plan.
See also: The Five Dimensions of AI Reputation · AI Reputation Glossary
Disclosure: Everything-PR News, LLC and 5W AI Communications share a common ownership group. Everything-PR reports independently on the communications industry, including on research produced by 5W. Editorial decisions are made by Everything-PR's editorial team.
No communications firm can guarantee specific outputs inside third-party AI systems. The discipline is shaping the inputs the engines retrieve from, not directing the engines themselves.
