Six AI Engines, Six Different Webs: Comparing Citation Source Overlap Across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode

Answer Capsule

ChatGPT, Gemini, Perplexity, Microsoft Copilot, Google AI Overviews, and Google AI Mode did not rely on one shared citation web across the 283 commercial markets in this study.

The six AI surfaces showed a clear hierarchy of source overlap.

At the Top-100 domain level, median Jaccard overlap ranged from 53.85% for Google AI Overviews versus Google AI Mode to 16.28% for Gemini versus Microsoft Copilot.

Because every eligible platform portfolio contained exactly 100 domains, those medians can be translated into approximate shared-domain counts:

• Google AI Overviews vs Google AI Mode: 70 shared domains
• Gemini vs either Google search surface: 47 shared domains
• Perplexity vs either Google search surface: 45 shared domains
• ChatGPT vs Perplexity or Copilot: 40 shared domains
• ChatGPT vs Gemini: 33 shared domains
• Gemini vs Copilot: 28 shared domains

The pair medians are only the first layer of the result.

Among the 260 markets with all 15 pairwise platform comparisons available, Google AI Overviews and Google AI Mode were the single highest-overlap pair in all 260 markets, with no ties.

Their advantage over the next-closest pair was not usually small. The median gap between Google AI Overviews/Mode and the runner-up pair was 20.06 percentage points. The middle 50% of markets had a lead between 15.67 and 24.13 points.

At the opposite end of the matrix, Microsoft Copilot was involved in the lowest-overlap pair in 227 of 260 complete markets, or 87.31%. It was part of the unique lowest pair in 185 markets, or 71.15%.

Within a typical complete market, the platform relationship was not mildly uneven. The median difference between the market's closest pair and furthest pair was 38.58 percentage points. The middle half of markets ranged from 34.33 to 43.66 points.

The market examples show how different those structures can be.

In Final Expense Insurance, Google AI Overviews and AI Mode shared 87 of their Top 100 domains, producing 76.99% Jaccard overlap. Gemini and Copilot shared only 24 domains, or 13.64%.

In Dental Implants, the Google search surfaces still led at 45.99%, but Google AI Overviews and Copilot shared only 7 domains, producing 3.63% overlap.

In Financial Technology and Banking Software, the Google sibling relationship was much less dominant. AI Overviews and AI Mode reached 39.86%, while ChatGPT and Google AI Mode were almost tied at 38.89%.

The practical conclusion is not that marketers need six unrelated content strategies. It is that one factual foundation can sit underneath six meaningfully different source ecosystems. Publisher outreach, PR targeting, citation monitoring, and competitive source analysis should therefore be measured separately by platform instead of assuming that authority on one AI surface transfers automatically to the others.

This is an observational study of cited registrable-domain portfolios. It does not establish platform architecture, training data, retrieval infrastructure, answer similarity, source quality, traffic, recommendation influence, or future ranking performance.

Key Findings

The Six Platforms Formed a Clear Citation-Overlap Hierarchy

Questions This Section Answers

1. Which AI platforms shared the most and least Top-100 citation sources across 283 commercial markets?
2. How many domains did the median ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode pair actually share?
3. Which AI platform pairs were routinely similar enough to share 30%, 40%, or 50% of their Top-100 source portfolios?

The full 15-pair matrix shows that the six AI surfaces did not cluster evenly.

The separation is substantial.

The Google search-surface pair was the only one with median overlap above 50%.

The next tier, Gemini with the Google search surfaces, was around 30.72%. Perplexity with those same surfaces was only slightly lower at 29.03%.

ChatGPT's highest median relationships were with Perplexity and Copilot at 25.00%.

Copilot's relationships with Gemini and the two Google search surfaces were the weakest pair medians in the matrix.

The table should not be interpreted as a ranking of product quality. It is a ranking of source-set resemblance.

A platform can produce a useful answer using a highly distinct source portfolio. Another platform can share many sources and still summarize them differently. Citation overlap measures one layer of the answer ecosystem, not the answer itself.

Jaccard Percentages Need to Be Translated Correctly

Questions This Section Answers

1. Does 25% Jaccard overlap mean two AI platforms share 25 of their Top 100 citation sources?
2. How many shared domains correspond to 53.85%, 30.72%, 25%, and 16.28% Jaccard overlap?
3. How should a CMO explain AI citation-source overlap without overstating platform fragmentation?

A Jaccard percentage is the size of the intersection divided by the size of the union.

For two lists containing exactly 100 domains each:

Jaccard = shared domains / all unique domains across both lists

That means 25% Jaccard does not mean 25 shared domains.

If two Top-100 lists share 40 domains, their union contains 160 domains:

40 / 160 = 25%

The same conversion applies across the matrix.

The corrected interpretation still shows large fragmentation.

At the ChatGPT-Gemini median, each Top-100 list contained roughly 67 domains absent from the other list.

At the Gemini-Copilot median, roughly 72 domains from each Top-100 list were absent from the other.

Even the closest pair, Google AI Overviews and Google AI Mode, had about 30 domains per list that were not shared at the median.

So “siblings” is a useful editorial analogy for the two Google search surfaces, but “identical twins” would be inaccurate even before looking at rank order or URLs.

Google AI Overviews and AI Mode Were the Closest Pair in Every Complete Market

Questions This Section Answers

1. Were Google AI Overviews and Google AI Mode the most similar citation-source pair only on average, or in individual markets too?
2. How consistently did Google AI Overviews and AI Mode outrank every other platform pair on source overlap?
3. How large was the typical citation-overlap lead between the Google search surfaces and the next-closest AI platform pair?

The aggregate 53.85% median is strong, but the market-level result is much more striking.

There were 260 markets in which all six platform portfolios were eligible and therefore all 15 platform-pair comparisons could be calculated.

In every one of those 260 markets, the highest-overlap pair was:

Google AI Overviews vs Google AI Mode

There were no ties for first place.

That means the Google sibling relationship was not created by a subset of unusually similar markets lifting the overall median. The pair ranked first in every complete market.

The magnitude of the lead varied.

In the median market, the Google search surfaces were therefore about 20 percentage points more similar to each other than the next-closest pair.

The largest separation appeared in Final Expense Insurance. Google AI Overviews and AI Mode reached 76.99% overlap, while the next-highest pair was only 35.14%. The lead was 41.86 points.

At the other extreme, Financial Technology and Banking Software produced a near-tie. Google AI Overviews and AI Mode reached 39.86%, while ChatGPT and Google AI Mode reached 38.89%. The lead was only 0.97 points.

This matters because it prevents an overly broad statement such as “Google AI Overviews and AI Mode always behave almost identically.”

They were always the most similar pair in the complete-market sample, but the absolute level of similarity ranged widely.

The dedicated analysis, Google AI Overviews and AI Mode Are Siblings, Not Twins. Gemini Is a Second Cousin. Perplexity Is the Family Friend Who Looks Suspiciously Related., develops that relationship in greater detail.

Gemini Was Usually Closer to the Google Search Surfaces Than Perplexity, but Not Always

Questions This Section Answers

1. Was Gemini consistently more similar to Google AI Overviews and AI Mode than Perplexity was?
2. How often did Perplexity resemble Google's AI search citation portfolios more than Gemini did?
3. Is the “Gemini second cousin, Perplexity suspiciously related family friend” analogy supported market by market?

The aggregate medians put Gemini slightly ahead of Perplexity in its relationship to the Google search surfaces:

• Gemini vs Google AI Overviews: 30.72%
• Gemini vs Google AI Mode: 30.72%
• Perplexity vs Google AI Overviews: 29.03%
• Perplexity vs Google AI Mode: 29.03%

That difference is real but modest.

The complete-market comparisons show that Gemini usually, but not universally, had the closer relationship.

For Google AI Overviews:

• Gemini had higher overlap than Perplexity in 155 of 260 markets;
• Perplexity had higher overlap in 95 markets;
• the two tied in 10 markets.

For Google AI Mode:

• Gemini had higher overlap in 150 of 260 markets;
• Perplexity had higher overlap in 89 markets;
• the two tied in 21 markets.

The median advantage for Gemini was small:

• about 0.98 percentage points against Google AI Overviews;
• about 1.57 points against Google AI Mode.

That is why the “family” analogy should be treated as a source-overlap metaphor, not a technical architecture claim.

Gemini generally sat closer to the Google search surfaces than Perplexity did. But in a large minority of markets, Perplexity was the more similar source ecosystem.

That variability matters for a CMO. Corporate ownership or product branding is not a sufficient substitute for measuring the actual citation portfolios in the market being targeted.

Copilot Was the Most Isolated Source Ecosystem in Most Complete Markets

Questions This Section Answers

1. Which AI platform was most often involved in the lowest-overlap citation pair across complete commercial markets?
2. How often was Microsoft Copilot the most source-divergent platform in the six-engine comparison?
3. Should a CMO treat Microsoft Copilot as a separate source-acquisition and publisher-measurement channel?

The bottom of the platform matrix is dominated by Copilot relationships.

Across the aggregate pair medians:

• Gemini vs Copilot: 16.28%
• Google AI Overviews vs Copilot: 16.96%
• Google AI Mode vs Copilot: 17.65%
• Perplexity vs Copilot: 19.76%
• ChatGPT vs Copilot: 25.00%

The complete-market analysis reinforces the pattern.

In 227 of 260 complete markets, 87.31%, Microsoft Copilot was involved in at least one pair tied for the lowest overlap in that market.

In 185 of 260 markets, 71.15%, Copilot was involved in the unique lowest-overlap pair.

The single most common bottom pair was Gemini vs Microsoft Copilot.

It was among the minimum-overlap pairs in 139 complete markets and was the unique minimum in 114.

The threshold distribution is similarly stark.

Copilot is therefore not simply “another member of the same source web” in the observed data.

That does not imply Copilot is worse. It implies its observed citation portfolio was frequently more distinct.

The dedicated pillar Microsoft Copilot Lives on a Different Web focuses on that pattern and the publisher implications it creates.

Google Search AI Was the Most Connected; Copilot Was the Least Connected

Questions This Section Answers

1. Which AI platform had the highest average citation-source similarity to the other five platforms?
2. Which AI platform occupied the most isolated position in the six-platform citation network?
3. How can CMOs compare overall platform connectedness without relying on one pairwise relationship?

Pairwise medians answer one relationship at a time.

A second descriptive view asks how similar each platform is to all five of the others within the same complete market.

For each of the 260 complete markets, we calculated each platform's mean Jaccard overlap with the other five platform portfolios. We then summarized those per-market averages across markets.

This is not an “authority score.” It does not say one platform is more important or better.

It is a descriptive measure of source-portfolio connectedness.

Google AI Mode and Google AI Overviews sat closest to the center of the observed six-platform source network because they were strongly connected to each other and moderately connected to Gemini and Perplexity.

Copilot sat furthest from the rest because several of its pair relationships were consistently low.

Perplexity's position is also notable. Its median connectedness exceeded Gemini's despite Gemini being closer to the two Google search surfaces on average. Perplexity compensated through relationships such as ChatGPT-Perplexity and its relatively balanced overlap across several platforms.

This is another reason not to infer the platform map from brand families alone.

The Typical Market Contained Both a Close Pair and a Very Distant Pair

Questions This Section Answers

1. How large was the gap between the most similar and least similar AI platform pairs inside the same market?
2. Do six AI citation ecosystems differ only modestly within a commercial category, or can the source relationships vary dramatically?
3. Which markets showed the widest and narrowest spread in platform-pair source overlap?

Within each of the 260 complete markets, we compared the highest and lowest of the 15 platform-pair overlaps.

The result was a wide internal spread.

That means a market can contain one platform pair sharing most of its leading domains and another pair sharing very little.

The largest observed spread was Final Expense Insurance:

• highest pair: Google AI Overviews vs AI Mode at 76.99%;
• lowest pair: Gemini vs Copilot at 13.64%;
• spread: 63.35 points.

Other markets with spreads above 50 points included:

• Invoice Factoring: 55.07 points
• Annuities: 54.87 points
• Gold IRAs and Precious Metals Dealers: 54.66 points
• Personal Loans and Online Lenders: 54.07 points
• Bad Credit Loans: 53.44 points
• Peer to Peer Lending: 52.38 points
• Herbal Supplements and Natural Remedies: 51.30 points
• Dermatologist Recommended Skin Care Brands: 51.00 points

The smallest spread among complete markets was Outdoor Retail and Gear Marketplaces at 16.64 points.

Even there, the closest pair was Google AI Mode vs AI Overviews at 31.58%, while the furthest was ChatGPT vs Copilot at 14.94%.

The practical implication is that “platform portability” is not a single uniform relationship. It is a summary across a network whose internal edges can differ dramatically.

Final Expense Insurance Shows the Google Sibling Effect at Its Strongest

Questions This Section Answers

1. How different can citation-source overlap be between platform pairs inside the same insurance market?
2. Which sources did Google AI Overviews and AI Mode share in Final Expense Insurance that other platform pairs did not?
3. What does a high Google sibling overlap look like compared with Gemini-Copilot fragmentation in the same market?

Final Expense Insurance contained 2,400 retained responses in the strict cohort and had 78.57% market-level persistence.

Its source portfolio was durable over time, but its platform relationships were highly uneven.

The Google search surfaces shared examples such as Aflac, Ethos, Choice Mutual, Fidelity Life, NerdWallet, Funeral Advantage, Colonial Penn, Policygenius, Mutual of Omaha, Progressive, Gerber Life, and MoneyGeek.

But even at 76.99% overlap, the lists were not identical.

Examples appearing in Google AI Mode's Top 100 but not Google AI Overviews' included GoHealth, Insurify, Experian, RiskQuoter, Banner Life, MassMutual, Ladder Life, and PolicyBazaar.

Examples appearing in Google AI Overviews but not AI Mode included TikTok, HelpAdvisor, Connelly Law, FreeAdvice, LinkedIn, and several final-expense-specific insurance domains.

The Gemini-Copilot pair shared only 24 domains. Their source universes therefore diverged far more sharply inside the same market.

This is the core “six different webs” phenomenon: the market itself is constant, but the source relationships depend heavily on which two AI platforms are compared.

Dental Implants Shows What Extreme Platform Fragmentation Looks Like

Questions This Section Answers

1. What does a very low-portability AI citation market look like across all six platforms?
2. How few Top-100 domains can two AI platforms share inside the same commercial market?
3. Can Google AI Overviews and AI Mode remain the closest pair even when most other platform relationships are extremely weak?

Dental Implants had one of the lowest market-level portability values in the complete cohort at 7.53%.

Yet the Google search surfaces still formed a distinct local cluster.

The top Google pair shared domains such as Aspen Dental, Affordable Dentures, CareCredit, ClearChoice, Nuvia Smiles, Reddit, YouTube, GoodRx, and several implant-focused dental sites.

But Google AI Overviews and Copilot shared only 7 domains out of two Top-100 portfolios.

That is not a small tactical difference. It is a fundamentally different source map.

A dental implant company using only Google AI Overviews citations to choose publishers or PR targets could therefore miss most of the domains appearing in Copilot's source portfolio.

The reverse is also true.

Credit Repair Shows What a More Portable Market Looks Like

Questions This Section Answers

1. Which commercial markets showed relatively high cross-platform citation-source agreement?
2. How different does a high-portability market look from Dental Implants or Life Insurance Companies?
3. Does higher portability eliminate the need for platform-specific source measurement?

Credit Repair provides a useful counterexample because its market-level portability was 29.87%, well above the 22.70% overall median.

Its pair matrix contained several relationships near or above 30%.

The lowest pair still shared 38 domains, compared with only 7 for the lowest Dental Implants pair.

That is a radically different cross-platform environment.

But even Credit Repair does not support a one-list-fits-all strategy. Its closest pair shared 72 domains. Its furthest pair shared 38. More than 60 domains in each Top-100 list could still be platform-specific at the low end of the matrix.

A market can therefore be relatively portable and still benefit from platform-specific measurement.

Pair Medians Hide Important Distribution Differences

Questions This Section Answers

1. How often did the major AI platform pairs reach at least 30%, 40%, or 50% source overlap?
2. Can two platform pairs have the same median overlap but different market-level distributions?
3. Why should a CMO look beyond one median platform-overlap statistic?

The 15-pair median table compresses hundreds of individual market comparisons into one number per pair.

Threshold counts reveal how differently those distributions behave.

The same lesson appears when two pairs share the same median.

ChatGPT-Perplexity and ChatGPT-Copilot both had a 25.00% median.

But in the 260 complete markets, ChatGPT-Perplexity had higher overlap than ChatGPT-Copilot in 130 markets, ChatGPT-Copilot was higher in 113, and they tied in 17.

An identical median does not imply identical market behavior.

This is why platform-pair rankings should be treated as distributions, not fixed constants.

A Single GEO Program Can Be Coordinated Without Pretending the Platforms Are the Same

Questions This Section Answers

1. Can a company run one GEO program across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode?
2. Which parts of AI search optimization can be shared across platforms, and which should be measured separately?
3. How should CMOs turn platform citation-overlap data into publisher, PR, and content decisions?

The evidence does not support six disconnected marketing organizations.

A company can and should maintain a shared foundation:

• one factual source of truth;
• consistent product and company information;
• strong first-party pages;
• structured data where appropriate;
• original research and evidence;
• clear authorship and attribution;
• consistent entity information;
• and coordinated brand positioning.

What should not be assumed to transfer automatically is the third-party source ecosystem.

The data suggests a practical operating model with three layers.

Shared source core

Identify domains that recur across several priority platforms in the same market.

These sources may offer efficient opportunities because improvements in factual coverage, editorial relationships, earned media, or brand visibility can potentially matter to several AI surfaces at once.

The word “potentially” matters. This study does not test placement effects.

Platform-specific source layer

Identify important domains that rank strongly on one or two platforms but are absent elsewhere.

Those sources are precisely what a global list or one-platform audit misses.

The How to Find the Market-Specific Publishers AI Systems Actually Cite framework provides a practical discovery workflow for this layer.

Measurement layer

Track each priority platform separately over time.

The parent Persistence-Portability Gap study found 63.93% median month-to-month persistence but only 22.70% median portability.

That is why source durability and source transfer belong in different fields on an executive dashboard.

The supporting article Can One GEO Strategy Work Across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode? develops this operating model further.

What This Means for Publishers

Questions This Section Answers

1. Can a publisher be an important AI citation source on one platform without being equally important on the others?
2. How should publishers describe platform-specific AI citation authority to advertisers?
3. What evidence should a publisher provide before claiming broad “AI search authority” across multiple engines?

The platform matrix creates a more precise language for publisher authority.

A publisher can have:

• market-specific authority, meaning it appears strongly within a commercial category;
• platform-specific authority, meaning it is prominent on one or several AI surfaces;
• cross-platform authority, meaning it recurs across several systems;
• persistent authority, meaning it remains part of the source portfolio over time.

Those are different assets.

A publisher that ranks strongly on ChatGPT but weakly on Copilot should not describe itself as universally strong across “AI search.”

But it also should not dismiss the ChatGPT footprint as unimportant merely because it does not transfer.

The correct advertiser-facing evidence package should identify:

• market;
• platform;
• collection period;
• response denominator;
• domain response presence;
• portfolio rank;
• persistence across completed periods;
• cross-platform coverage;
• cited URLs;
• and commercial disclosure.

The article How Publishers Should Think About Platform-Specific AI Citation Demand focuses on how publishers can package that distinction without promising outcomes the data cannot support.

Methodology

Questions This Section Answers

1. How were cross-platform AI citation-source overlaps calculated across the 283-market benchmark?
2. Why do the 15 platform pairs have different market counts in the comparison table?
3. What does a complete six-platform market mean in this study?

The primary benchmark covers completed observations from July through September 2026 across six consumer AI surfaces:

• ChatGPT
• Gemini
• Perplexity
• Microsoft Copilot
• Google AI Overviews
• Google AI Mode

The strict qualifying cohort contains:

• 283 current tracked markets;
• 876,605 retained AI responses;
• 8,718,629 citation occurrences resolved to registrable domains;
• 6,825,821 response-domain appearances;
• 308,807 distinct registrable domains.

Platform portfolios

Within each market-platform cell, completed months were pooled and registrable domains were ranked by the number of distinct retained responses in which the domain appeared.

Repeated citations to the same domain inside one response counted once for ranking purposes.

An eligible platform portfolio required:

• at least 100 retained responses;
• at least 100 resolved registrable domains;
• an actual ranked list of exactly 100 domains.

Thin or missing cells were excluded, never scored as zero.

Pair overlap

For each eligible platform pair within a market, Top-100 domain sets were compared using Jaccard similarity:

intersection / union

The platform-pair table reports the median across all eligible markets for that pair.

Because eligibility varies by platform, pair support ranges from 265 to 283 markets.

Complete six-platform subset

A market enters the complete six-platform subset when all six platform portfolios are eligible, which allows all 15 pairwise comparisons to be calculated.

There were 260 such markets for this platform-pair analysis.

That subset is distinct from the 246-market all-three-month/all-six-platform marginal subset used in some persistence sensitivity analyses. The two numbers answer different support questions and should not be conflated.

Connectedness calculation

For the descriptive platform-connectedness table, each platform's five pair overlaps were averaged within each complete market. Those market-level averages were then summarized across the 260 complete markets.

This is not a causal metric or authority score. It is simply a compact description of how closely each platform's source portfolio tended to resemble the other five.

Important Limitations

Citation overlap is not technical architecture

The study cannot determine whether two platforms share retrieval infrastructure, licensing agreements, model training data, ranking systems, indexes, or source-selection logic.

The Google sibling result is a citation-source resemblance finding only.

Domain overlap is not URL overlap

The unit is the registrable domain.

Two platforms can both cite the same domain while citing different pages, authors, products, or arguments.

Set overlap ignores rank order

A domain ranked #2 on one platform and #92 on another counts as shared in the Top-100 Jaccard calculation.

Top-25 and Top-50 sensitivity analyses provide additional depth context, but this article's primary platform matrix uses Top 100.

The historical collection is not a certified fixed-prompt experiment

The underlying data does not establish an identical panel of prompts, prompt versions, model versions, countries, retrieval settings, or collection conditions across every platform and month.

The results describe observed source portfolios, not controlled platform effects.

Citation is not recommendation

A domain can be cited without the AI system recommending the domain's company, advertiser, or products.

Likewise, a company can be recommended while a third-party publisher is cited as evidence.

Citation authority and recommendation authority require separate measurement.

Platform distinctiveness is not commercial value by itself

A highly distinctive source portfolio can create unique publisher opportunities, but it does not establish audience size, conversion value, customer intent, media efficiency, or return on spend.

Those require separate business evidence.

Conclusion

The six AI surfaces in this benchmark did not operate on one interchangeable citation web.

They formed a structured but fragmented source network.

Google AI Overviews and Google AI Mode were the clear exception. Their median Top-100 overlap was 53.85%, equivalent to 70 shared domains between two equal 100-domain lists. More importantly, they were the highest-overlap pair in all 260 markets with complete six-platform support.

But even that relationship was not uniform. The sibling pair ranged from 29.03% to 76.99% across complete markets, and its lead over the next-best pair ranged from 0.97 to 41.86 percentage points.

Outside the Google search surfaces, overlap dropped sharply.

Gemini and the Google search surfaces centered around 30.72%. Perplexity and those same surfaces centered around 29.03%. ChatGPT's strongest median relationships were 25.00%. Gemini and Copilot fell to 16.28%.

Copilot was involved in the lowest-overlap pair in 87.31% of complete markets, illustrating how distinct one platform's source ecosystem can become even when the commercial question remains the same.

The median complete market contained a 38.58-point spread between its closest and furthest platform pairs.

That is the evidence behind the title “Six AI Engines, Six Different Webs.”

It does not mean there is no shared web. There is.

It means the shared core sits inside much larger platform-specific source portfolios.

For CMOs, the practical implication is to coordinate brand facts, content quality, and editorial standards centrally while measuring third-party source authority separately on each priority AI surface.

For publishers, it means platform-specific citation authority can be a real asset, but it should be described precisely rather than converted into a universal “AI authority” claim.

For researchers, it means cross-platform AI citation analysis should preserve the full pair matrix. One portability average cannot capture a network where one pair shares 70 of its Top 100 sources and another shares only 28.

Related LLM Authority Index Research

• The Persistence-Portability Gap at Scale: 282 of 283 Markets Show Higher AI Citation Persistence Than Cross-Platform Portability Across 8.72 Million Citation Occurrences provides the parent market-level benchmark.
• Persistence vs Portability: Why a Source That Wins on ChatGPT May Not Transfer to Gemini or Google AI examines source transfer from ChatGPT in detail.
• Can One GEO Strategy Work Across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode? translates the six-platform matrix into program design.
• Which AI Platforms Share the Most Citation Sources, and Which Diverge the Most? focuses on pair ranking and extremes.
• Google AI Overviews and AI Mode Are Siblings, Not Twins. Gemini Is a Second Cousin. Perplexity Is the Family Friend Who Looks Suspiciously Related. focuses on the Google-family relationship.
• How Publishers Should Think About Platform-Specific AI Citation Demand focuses on publisher-side implications.
• Microsoft Copilot Lives on a Different Web examines Copilot-specific divergence.
• AI Search Has a Tiny Universal Core and a Massive Specialist Long Tail examines domain breadth across markets and platforms.
• How to Find the Market-Specific Publishers AI Systems Actually Cite provides the source-discovery workflow used for publisher diligence.
• The 2026 AI Commercial Citation Authority Index separates breadth, depth, and platform reach.

The earlier 2026 AI Citation Authority Study and cross-platform citation-overlap study use a separate earlier research population and should not be combined numerically with this benchmark.

About LLM Authority Index

LLM Authority Index measures how AI systems cite sources, surface publishers, and structure authority across commercial markets. Its research separates citation depth, market breadth, platform breadth, persistence, portability, and recommendation authority so that source overlap is not mistaken for platform equivalence or recommendation influence.