Why AI Engines Cite Deloitte: Source Authority in the Machine Relations Index

Deloitte.com's citation volume fell 62% in the latest Machine Relations Index measurement cycle — from 148 citations to 56. Its MRI consensus score dropped from 80.3 to 76, falling below 80 for the first time. Google AI Mode, previously Deloitte's dominant citing engine at 33.8% share, contracted 84% from 50 to 8 citations. Claude contracted 86%, from 42 to 6. This analysis uses current Machine Relations Index data to examine the steepest single-cycle analyst-source decline in MRI measurement history: what collapsed, what held, and what the redistribution pattern reveals for operators building citation authority.

Last updated: June 27, 2026

Deloitte MRI Profile: 56 Citations Across 6 AI Engines #

The Machine Relations Index measures source citation authority across AI answer engines using a composite methodology (MRI Score v1.1, 6-engine). Deloitte's current profile shows a source that dropped below 80 consensus for the first time while retaining Elite tier status.

MRI consensus score: 76 (Elite tier, B-confidence)

Deloitte ranks #2 among 229 analyst and consulting research sources tracked in the MRI, placing it at the 99.6th percentile within that source role. The analyst source field itself contracted from 295 to 229 tracked domains — a 22% reduction in the measured competitive set. Deloitte's weighted authority score of 33.9 reflects a collapse from 75.1, a 54.9% decline that is the steepest weighted-authority contraction for any top-5 analyst source in MRI history. The measurement covers 6,077 total domains and 17,821 source events.

Every metric declined except engine breadth (still cited by all 6) and position quality (which improved mechanically as low-quality citations dropped off — the same paradox documented in the Crunchbase MRI analysis). Temporal consistency lost its perfect 10.0 score despite maintaining 22 cited days, reflecting a change in measurement period normalization. Confidence downgraded from A to B as the citation sample shrank below the threshold for reliable pattern establishment.

Citation Distribution by AI Engine #

The 30-day engine breakdown reveals a complete hierarchy reshuffling — the most dramatic redistribution in any single MRI cycle for analyst research sources.

Google AI Mode and Claude each collapsed over 84% — together losing 78 citations that account for 85% of Deloitte's total volume decline. In the prior cycle, these two engines produced 62.2% of all Deloitte citations. They now produce 25%.

Perplexity became Deloitte's top citing engine — not by growing, but by declining the least among the previously dominant engines. Its 55.6% contraction (36 to 16) is severe, but relative to Google AI Mode's 84% and Claude's 86%, it reads as stability. Perplexity's share grew from 24.3% to 28.6%.

Gemini is the only engine that increased Deloitte citations in absolute terms, rising from 12 to 14 (+16.7%). Google AI Overviews grew from 1 to 6 (+500%), making it tied for third with Claude and ChatGPT. Research on Google AI Overviews source selection finds that pages with 15 or more Knowledge Graph entities show 4.8x higher selection probability — a structural test Deloitte's research pages are positioned to pass.

The distribution is now radically flat. The spread between Deloitte's top engine (Perplexity, 16) and bottom engines (Claude/ChatGPT/AI Overviews, 6 each) is just 10 citations. In the prior cycle, the spread was 49 (Google AI Mode at 50 vs. AI Overviews at 1). This flattening occurred through collapse of the dominant engines, not growth of smaller ones — the same pattern observed in Crunchbase's citation redistribution and consistent with broader research on AI engine citation divergence.

The Google AI Mode collapse is particularly significant for analyst sources. In the prior measurement, Google AI Mode was the dominant retrieval engine for enterprise consulting content, accounting for 33.8% of Deloitte's citations. Its 84% contraction aligns with broader structural changes: Google I/O 2026 introduced changes to how AI Mode cites brands, and the overlap between top-10 organic rankings and AI citation selection collapsed from 75% to between 17% and 38% by early 2026, decoupling traditional search authority from AI retrieval. The Crunchbase analysis documented a simultaneous 52.5% Google AI Mode contraction for market databases — Deloitte's 84% drop indicates analyst sources were hit substantially harder than data sources in the same engine update.

Claude's 85.7% collapse (42 to 6) is the steepest single-engine decline in Deloitte's measurement history. Claude went from Deloitte's second-largest citing engine to tied for last. This contraction is disproportionate compared to Claude's decline for other source types — Claude remained Crunchbase's top citing engine despite a 19.2% decline — suggesting that Claude's retrieval updates specifically deprioritized analyst/consulting research relative to structured data sources. An analysis of AI citations and Google organic visibility changes found that AI citation declines closely mirror organic traffic drops, with an average citation decline of 22.5% across studied sites — Deloitte's 86% Claude contraction far exceeds that baseline, pointing to source-type-specific retrieval changes beyond general visibility trends.

What Makes Deloitte Citation-Eligible #

Deloitte's structural properties — the features that make it retrievable by AI engines — remain intact despite the volume collapse. The decline is in how frequently engines choose to retrieve it, not in whether they can.

Analyst frameworks with extractable structure #

Deloitte publishes enterprise technology research in formats that answer engines can parse: survey reports with statistical findings, maturity frameworks with defined stages, and benchmarking analyses with named vendors and measurable outcomes. The State of AI in the Enterprise 2026 report surveyed 3,235 business and IT leaders across 24 countries and found that 85% of companies expect to customize AI agents to fit their business needs, while only 21% have mature agentic AI governance. Deloitte's AI agents scaling research further documents how autonomous agents are transforming enterprise operations. These are exactly the kind of structured, findings-first publications that AI retrieval systems favor.

The Authority Signals Framework, which analyzed 10,038 citations across 542 sources, identifies structural clarity, factual density, and methodological transparency as dominant predictors of citation selection. Deloitte's research pages satisfy these criteria. But citation eligibility is a threshold, not a guarantee — above that threshold, competitive dynamics and retrieval stack updates determine volume. Deloitte's volume collapsed while its structural properties held.

Query coverage across enterprise verticals #

Deloitte's cited queries contracted from 45 to 29, and its vertical coverage dropped from 9 to 8. The verticals represented in the current cycle are enterprise AI, fintech, healthtech, HR tech, infrastructure/devtools, and legal/compliance. Current sample queries from the MRI measurement include:

• "AI chip and hardware infrastructure for enterprise compute"
• "AI governance and compliance frameworks for enterprise"
• "AI infrastructure companies entering enterprise market"
• "B2B payment infrastructure modernization and market shifts"
• "ESG reporting requirements and enterprise compliance challenges"

The remaining queries reflect the same enterprise technology core: governance, infrastructure, regulation, and compliance. Deloitte lost 16 queries — a 35.6% contraction in query diversity — indicating that AI engines are no longer defaulting to Deloitte for peripheral queries where alternative analyst sources exist.

Temporal consistency decline #

Deloitte's temporal consistency score fell from a perfect 10.0 to 7.6. The source was cited on 22 measured days — the same count as the prior cycle — but the component score dropped, reflecting a broader measurement window. The loss of perfect temporal consistency means Deloitte is no longer a guaranteed daily answer for enterprise technology queries. There are now days when no AI engine retrieves a Deloitte page in response to any of its previously cited query categories.

This is structurally different from the prior cycle, where Deloitte appeared in at least one AI answer every single measured day. Research on how AI engines evaluate source trust across industries shows that temporal consistency is one of the strongest indicators of default retrieval status. Losing it signals that Deloitte is transitioning from default-position status to competitive-position status in the AI retrieval stack.

Source Role: Analyst Research Competitive Landscape #

Among 229 tracked analyst and consulting research sources (down from 295), Deloitte retained its #2 position at the 99.6th percentile. The field contraction — 66 fewer tracked analyst domains — reflects both the broader index contraction (6,077 total domains, down from 6,804) and the elimination of low-citation analyst sources from the tracked set.

Deloitte's 4.3-point consensus decline, combined with the category contraction, raises a structural question: are AI engines becoming less dependent on analyst research sources as their own evaluative capabilities improve? The volume decline is not Deloitte-specific — the entire analyst research category contracted — but Deloitte's 62% decline is disproportionate to its field position, suggesting the largest analyst sources faced the steepest retrieval stack adjustments.

Analyst research sources occupy a distinct position in the AI citation stack compared to market databases. Where market databases like Crunchbase provide structured entity data that satisfies factual queries, analyst sources provide frameworks, assessments, and projections that satisfy evaluative queries. An enterprise buyer asking "AI governance frameworks for enterprise" needs an analyst assessment; an enterprise buyer asking "Series B funding rounds in HR tech" needs a data source. The 62% volume decline suggests that AI engines are finding more ways to satisfy evaluative queries without defaulting to a single analyst source.

The Deloitte Paradox: Citation Authority Despite Fabrication #

In late 2025, Deloitte Canada was found to have used AI-fabricated citations in policy submissions to the Canadian government. The incident involved AI-generated references that did not correspond to real publications — the exact kind of reference hallucination that undermines source trust.

Deloitte's MRI score declined from 80.3 to 76 this cycle, but the decline correlates with retrieval stack changes across all engines, not with the fabrication incident. The fabrication affected Canadian policy documents; the citation decline spans enterprise technology queries globally and tracks with similar — though less severe — declines for sources like Crunchbase and G2 that had no fabrication incidents.

This reinforces the finding from the prior analysis: AI citation authority operates at the page level and structural level, not at the organizational reputation level. The 62% citation collapse has the same mechanical cause as Crunchbase's 28% decline — retrieval stack updates within Google AI Mode and other engines — not a trust penalty from the fabrication incident. Research on measuring AI overview source quality confirms that source selection in generative search operates primarily at the URL level, with limited propagation of trust or distrust across a domain. A meta-analysis of six 2026 studies covering 366,087 real-world citations across three AI providers confirms that citation selection is driven by structural content properties — entity density, data attribution, chunk clarity — not organizational brand signals. An analysis of citation attribution in LLM deep research agents found that citation verification remains inconsistent even in sophisticated retrieval systems, reinforcing the page-level rather than domain-level nature of source trust.

What Operators Can Learn from Deloitte's Citation Collapse #

1. Analyst sources are more vulnerable to retrieval stack changes than data sources. Deloitte's 62% citation decline is more than double Crunchbase's 28% decline in the same measurement cycle. Market databases provide factual answers that retrieval systems need; analyst sources provide evaluative answers that retrieval systems can increasingly generate independently. As AI engines improve their ability to synthesize assessments from multiple sources, the default retrieval of any single analyst source becomes less necessary.

2. An 84% single-engine contraction redefines concentration risk. Deloitte's Google AI Mode citations fell from 50 to 8. Any source deriving more than 30% of citations from a single engine faces catastrophic risk if that engine's retrieval stack deprioritizes their source type. Deloitte's prior 33.8% Google AI Mode share was a concentration warning. The 84% contraction converted that warning into a realized loss of 42 citations — more than Deloitte's entire current ChatGPT, Claude, and AI Overviews citation volume combined.

3. Weighted authority is the unmasked signal. Deloitte's consensus score fell 4.3 points (80.3 to 76). Its weighted authority fell 55% (75.1 to 33.9). Consensus masks severity because it weights dimensions like engine breadth (unchanged at 40/40) and position quality (improved mechanically from 1.7 to 2.2) alongside volume-dependent metrics. Weighted authority amplifies volume and position quality together, revealing the true magnitude. Track weighted authority as the lead indicator; treat consensus as the lagging composite.

4. Position quality improves when low-quality citations drop first. Deloitte's average citation position improved from 8.8 to 6.8, and its position quality score rose from 1.7 to 2.2. This happened because the citations that disappeared were the ones in positions 10+ — the end of answer lists. The remaining 56 citations concentrate in higher positions. This is contraction masquerading as precision, the same pattern documented in the Crunchbase analysis. Position quality improving during a volume collapse is not a positive signal.

5. Query diversity contraction reveals the competitive frontier. Deloitte lost 16 queries (45 to 29, -35.6%). The queries that survived are the ones where Deloitte has the deepest, most structured content — governance frameworks, infrastructure assessments, compliance evaluations. The queries that dropped are likely ones where competing sources now satisfy the retrieval system more directly. The competitive frontier for analyst citation authority is moving toward specificity: the sources that retain citations will be the ones with the most extractable, attributed data per query, not the broadest coverage.

6. Google AI Overviews growth is the structural counterweight. Deloitte's AI Overviews citations grew from 1 to 6, a 500% increase. AI Overviews appears directly in Google search results — the highest-traffic AI surface. While 6 citations cannot offset the 92 lost across other engines, AI Overviews is the only surface where Deloitte's analyst content gained traction this cycle. For operators, AI Overviews citation share may be the most strategically important engine-specific metric because it reaches users at the point of search intent, not through a separate AI application.

How This Connects to Machine Relations #

Deloitte's 62% citation collapse is the most significant analyst-source decline in MRI measurement history. It exposes a structural question for the Machine Relations framework: are evaluative sources — analyst research, consulting assessments, framework publications — entering a structural decline in AI retrieval, or is this a cyclical contraction driven by engine updates?

The evidence points toward both. The Google AI Mode and Claude contractions are engine-specific events — retrieval stack updates that deprioritized analyst content. But the magnitude of those contractions (84% and 86%) compared to market database contractions (52.5% and 19.2% for Crunchbase in the same engines) suggests that evaluative content is structurally more substitutable than factual data. An AI engine can synthesize an assessment from multiple analyst sources; it cannot synthesize a funding round from nothing.

For practitioners building citation architecture, Deloitte's trajectory offers a category-defining lesson: source authority in AI engines is not a permanent position. It is a continuously contested resource that can contract 62% in a single measurement cycle. The properties that maintain citation eligibility — structured data, extractable claims, cross-engine retrievability — are necessary but no longer sufficient to maintain volume. Volume defense requires being the single most direct, most attributed, most extractable answer for each query the source targets.

Deloitte remains Elite-tier at the 99.6th percentile. Its structural properties are intact. But the gap between structural integrity and citation volume is wider than at any point in MRI measurement history. The trajectory from 148 to 56 is not a fluctuation — it is the retrieval stack rewriting the rules of analyst citation authority.

FAQ #

What is Deloitte's MRI score? #

Deloitte.com has a Machine Relations Index consensus score of 76, placing it in the Elite tier with B-confidence. It ranks #2 among 229 analyst and consulting research sources tracked in the MRI, with 56 citations across 6 AI engines over a 30-day measurement period. The score dropped from 80.3, falling below 80 for the first time. Weighted authority collapsed 55% from 75.1 to 33.9 — the steepest single-cycle decline for any top-5 analyst source.

Which AI engines cite Deloitte most? #

Perplexity now leads with 28.6% of Deloitte's 30-day citations (16 of 56), followed by Gemini at 25% (14), Google AI Mode at 14.3% (8), and Claude, ChatGPT, and Google AI Overviews tied at 10.7% each (6 apiece). The engine hierarchy completely reshuffled: Google AI Mode fell from #1 (50 citations, 33.8% share) to #3 (8 citations, 14.3%), and Claude fell from #2 (42 citations, 28.4%) to tied-last (6 citations, 10.7%).

Why did Deloitte's citations fall 62%? #

The primary cause is retrieval stack updates within Google AI Mode (-84%) and Claude (-86%), which together lost 78 citations accounting for 85% of Deloitte's total volume decline. These contractions coincided with similar declines for other Elite-tier sources like Crunchbase (-28%), suggesting cross-source engine updates. However, the disproportionate severity for analyst sources versus market databases indicates evaluative content is more vulnerable to retrieval stack changes than factual data sources.

Is Deloitte losing AI citation authority? #

Yes. Deloitte's citation volume fell 62% in the latest cycle — the steepest decline for any Elite-tier analyst source in MRI history. Its consensus score dropped 4.3 points, confidence downgraded from A to B, weighted authority collapsed 55%, query diversity contracted 36%, and temporal consistency lost its perfect score. Deloitte retains Elite status and the #2 analyst research ranking, but the trajectory is unambiguously declining.

How is the Machine Relations Index calculated? #

The MRI (v1.1, 6-engine) measures citation authority across Perplexity, ChatGPT, Gemini, Claude, Google AI Mode, and Google AI Overviews. The consensus score combines five components: engine breadth (how many engines cite the source), query diversity (how many distinct queries trigger citations), vertical spread (industry coverage), position quality (where the source appears in citation lists), and temporal consistency (how many days the source is cited). The index currently tracks 6,077 domains across 17,821 source events. For methodology details, see What is Share of Citation.