How AI Search Is Recommending Crypto Exchanges: Monthly Trends

Executive Summary

Kraken holds the category lead in October 2026 with 78.2% valid recommendation coverage, ahead of the next brand, Binance, by 15.6 points (78.2% versus 62.6%). The gap to second position widened this month: Kraken rose 14.3 points from September (63.9% to 78.2%) while Binance rose 14.5 points (48.1% to 62.6%), keeping the two brands separated by more than fifteen points.

Six of the nine tracked brands recorded significant gains from the May 2026 baseline, but the sharper story this month is the two-month climb that followed August. Kraken posted the largest baseline-to-current gain among the six, up 37.0 points (41.2% to 78.2%). Crypto.com followed with a 33.8-point gain (24.0% to 57.8%), Binance rose 33.4 points (29.2% to 62.6%) in its second consecutive month of significant improvement, and Gemini rose 28.6 points (21.6% to 50.2%). Uniswap rose 9.5 points (3.9% to 13.4%) and Uphold rose 13.0 points (5.1% to 18.1%).

The sharpest decliner is Robinhood, down 19.6 points across the baseline-to-current window (19.6% to 0.0%), with no presence in the October qualified set. PayPal likewise fell to 0.0% from 1.2% in May, and Pionex.US sits at 0.1%, down 0.6 points from May, though its prior-to-current move edged up 0.1 point.

Against the full series the category sits at its strongest measured level since the benchmark began recording. October's 78.2% leader reading is the highest of the five months, above the July peak of 72.4% and well clear of the August trough of 17.6%. Most brands with meaningful presence rose on both the baseline and the prior-to-current comparison this month; Uniswap and Uphold were the exceptions on the prior-to-current measure, posting essentially flat month-over-month readings. The field has not merged into a single cluster: the spread between first and fourth place is 28.0 points.

Each monthly run begins with 800 prompt-surface observations across the benchmark's defined AI/search surface universe. In October 2026, those 800 prompts covered 569 unique questions; 800 mentioned a tracked brand or competitor; 750 were relevant and 50 were irrelevant. The public metrics use the 673 qualified observations that survive both qualification stages. In the May 2026 baseline, no prompt funnel was recorded and 1,591 qualified observations were published; July 2026 ran from 800 prompts with 703 qualified, August 2026 from 800 prompts with 681 qualified, and September 2026 from 800 prompts with 676 qualified.

Key Findings

AI Response Inconsistency Alerts

Questions This Section Answers

• Which exchanges had factual conflicts flagged in AI platform responses this month?
• What availability and fee claims did platforms disagree on for Binance and Kraken?

Five critical or high severity factual inconsistencies were detected this month, spanning five AI platforms: ChatGPT, Copilot, Gemini, AI Mode, and Perplexity. The conflicts concentrate in two companies, Binance and Kraken, and both centre on facts a buyer would rely on before opening an account: whether an exchange is actually available in the United States, and what it charges.

Binance

Two inconsistencies were flagged for Binance, both about availability.

Critical (confidence 0.95), availability for U.S. users. When asked "How to buy Tao Bittensor in the USA?", Gemini stated that Binance restricts U.S. users or operates separate localized entities, and that Binance.com is generally unavailable or restricted for U.S. IP addresses and KYC. Perplexity, answering the same question, stated that Binance is a U.S.-accessible exchange that supports TAO and can be used to buy TAO in the USA. Gemini cited a Cryptonews buying guide, Changelly's TAO page, and a Finder guide; Perplexity cited Kraken's TAO page, a Crypto.com learn article, and Binance's own "how to buy Bittensor" page.

High (confidence 0.9), XDC availability on Binance. When asked "Why can't I buy an XDC Network?", AI Mode stated that XDC is not listed on standard Binance but is available on Binance.US, while Copilot stated that XDC is not listed on Binance in the U.S. AI Mode referenced a Binance.US support article announcing the XDC listing, a Reddit thread, and xdc.org; Copilot referenced FinanceFeeds, Baltex, and Securities.io. The flagged Binance.US support page ("We're excited to announce that XDC is now listed on Binance.US") was scored as supporting the AI Mode side of the conflict.

Kraken

Three inconsistencies were flagged for Kraken, one on pricing and two on availability.

High (confidence 0.9), availability in New York. When asked "What crypto exchanges are legal to US citizens?", Gemini described Kraken as FinCEN-registered and a major legal U.S. exchange without noting New York restrictions, while Copilot stated that Kraken is available in 48 states and not available in NY (and WA). Gemini cited Changelly and Coin Bureau; Copilot cited ChainChoice, Datawallet, and BeginnerCrypto. Two flagged sources, ChainChoice ("NY-restricted") and BeginnerCrypto ("Available in: 48 states (not NY, WA)"), were scored as supporting the Copilot side.

High (confidence 0.9), availability in New York. When asked "Which is the best crypto exchange site?", Gemini presented Kraken as available and recommended for active traders with no mention of New York restrictions, while Copilot stated Kraken is not available in Maine or New York. Gemini cited Changelly and Newhedge; Copilot cited Invezz, Yahoo Finance, and Datawallet. The flagged Yahoo Finance page ("Not available in Maine or New York") supported the Copilot side, while a flagged Tradingpedia excerpt ("Kraken offers the lowest fees among major U.S.-regulated exchanges") supported the Gemini side.

High (confidence 0.9), Kraken Pro maker/taker fees. When asked "Who has cheaper fees than Coinbase?", Gemini stated Kraken Pro fees drop to a maximum of 0.16% maker and 0.26% taker, while ChatGPT stated Kraken Pro starts at 0.25% maker and 0.40% taker. Gemini cited Paybis, Datawallet, and Coinsutra; ChatGPT cited Coinbase's own advanced trade fee page, Kraken's "lowest fee crypto exchange" learn page, and Gemini's Cryptopedia. Two flagged sources, Coinsutra ("Kraken Maker Fee 0.16%, Taker Fee 0.26%") and CoinLedger ("Kraken Pro 0.16% / 0.26%"), were scored as supporting the Gemini side.

The pattern worth noting commercially: these are not stylistic differences. Availability and fee accuracy are threshold facts. Where AI platforms disagree on whether an exchange serves a U.S. state or what it charges, the buyer receives a conditional answer and may resolve the conflict by choosing a different option entirely.

Benchmark Context

Questions This Section Answers

• How does the qualified benchmark set differ from the raw prompt collection?
• What were the benchmark-level changes in recommendation output between May and October 2026?

The report separates the raw collection universe from the qualified analysis set. Brand-level recommendation percentages are calculated within the qualified benchmark set.

Intermediate months: July 2026 ran 800 prompts with 703 qualified, August 2026 ran 800 prompts with 681 qualified, and September 2026 ran 800 prompts with 676 qualified. The May 2026 measurement did not record a prompt funnel, so only the 1,591 qualified observations are available for that month.

Benchmark-Level Metrics

August 2026 remains the series low point, with recommendation-shaped answer share at 10.0% and valid recommendation shortlist share at 17.6%. September recovered to 30.3% and 61.2% respectively. October pushed both to their highest recorded levels, at 51.7% and 79.0%, the strongest recommendation-oriented output the benchmark has measured for this vertical.

AI Recommendation Trend

Questions This Section Answers

• Which brands led in recommendation coverage in October 2026?
• Which brands posted significant gains from the May baseline?

Leader Extends Its Advantage as the Field Climbs Together

Most brands that recorded meaningful presence moved by more than normal month-to-month variation this month. The category-level gain is a combination of several individually significant rises rather than one brand's outlier movement; the table shows the distribution across brands, not the mechanism behind it. Two brands' prior-to-current movement stayed inside normal variation: Uniswap, up 1.9 points from September (11.5% to 13.4%), and Uphold, down 0.2 points from September (18.3% to 18.1%).

What Changed This Month

Questions This Section Answers

• How did Kraken and Binance differ on coverage gain versus rank-one placement this month?
• Why are Uniswap and Uphold's month-over-month changes read as directional rather than established?
• What does it mean for PayPal and Robinhood to have no presence in the October qualified set?

Kraken

Kraken's valid recommendation coverage rose from 41.2% in May to 78.2% in October, up 37.0 points. The prior-to-current move is also significant: up 14.3 points from 63.9% in September, extending the climb to a second consecutive month.

The gain reached placement depth. Kraken's recommended top-three rate rose from 30.9% in May to 72.7% in October, up 41.8 points, and its rank-one rate rose from 15.0% to 32.5%, up 17.5 points. Raw mention presence climbed from 48.3% to 98.1%, up 49.8 points. In absolute terms, Kraken earned 526 valid recommendations in October out of 673 observations, compared with 432 out of 676 in September and 120 out of 681 in August.

The distinction to notice: Kraken is now the default answer for a third of qualified observations, not merely a frequent one. A 32.5% rank-one rate means roughly one in three recommendation-shaped prompts returns Kraken as the top pick. That is a different commercial position from broad-but-shallow presence, and it is the kind of placement that is hardest for a competitor to displace quickly.

Highest-priority diagnostic: Which prompt clusters and surfaces produce Kraken's rank-one placements, and does the lead depend on a small set of surfaces that a competitor could contest first?

Binance

Binance's coverage rose from 29.2% in May to 62.6% in October, up 33.4 points. The prior-to-current gain was also significant at 14.5 points (48.1% to 62.6%), its second straight monthly rise.

Placement improved across the board. Top-three rate rose from 17.0% in May to 42.8% in October, up 25.8 points, and rank-one rate rose from 5.1% to 9.8%, up 4.7 points. Raw mention presence moved from 39.0% to 85.6%, up 46.6 points. Binance held 421 valid recommendations in October versus 325 in September.

The distinction: Binance is now second on coverage and second on top-three placement, but its 9.8% rank-one rate is less than a third of Kraken's 32.5%. The brand is winning inclusion and near-the-top placement without yet winning the lead recommendation at comparable frequency.

Highest-priority diagnostic: Which prompts place Binance in the shortlist but behind another brand at rank one, and which competitor occupies the top slot in those answers?

Crypto.com and Gemini

Crypto.com rose from 24.0% in May to 57.8% in October, up 33.8 points, with a 9.6-point prior-to-current gain (48.2% to 57.8%). Gemini rose from 21.6% to 50.2%, up 28.6 points, with a 7.6-point current-month gain (42.6% to 50.2%).

In absolute terms, Crypto.com held 389 valid recommendations in October versus 326 in September, and Gemini held 338 versus 288. Top-three rates rose for both: Crypto.com from 6.9% to 18.3%, Gemini from 5.5% to 19.2%. Neither brand moved meaningfully on rank-one placement, with Crypto.com at 0.7% and Gemini at 0.4%.

The distinction: Crypto.com and Gemini now sit within 7.6 points of each other on coverage and within 1.0 point on top-three rate, but both are far behind on rank-one. They are consistently recommended and rarely the recommendation.

Highest-priority diagnostic: Which prompt clusters return Crypto.com or Gemini at rank two or three, and which brand holds rank one in those same answers?

Uniswap and Uphold

Uniswap rose from 3.9% in May to 13.4% in October, up 9.5 points, and Uphold from 5.1% to 18.1%, up 13.0 points. Both remain well below the leading four.

Uniswap's prior-to-current move was 1.9 points, inside normal variation, while Uphold's was 0.2 points down, also inside normal variation. Uniswap held 90 valid recommendations in October versus 78 in September, and Uphold held 122 versus 124. Top-three rates are low for both: Uphold at 4.3%, Uniswap at 2.4%. Rank-one placement is negligible, at 0.4% and 0.3% respectively.

The distinction: both brands are visible and occasionally recommended, but their coverage is not translating into shortlist depth. The small absolute counts matter here: a swing of a handful of valid recommendations moves the Uphold and Uniswap percentages by a full point or more, so month-to-month changes for these two should be read as directional rather than established.

Highest-priority diagnostic: Which comparison or consideration prompts include Uniswap and Uphold in the recommended set, and do those prompts convert at all to top-three placement?

PayPal, Robinhood, and Pionex.US

PayPal's coverage fell from 1.2% in May to 0.0% in October, down 1.2 points. Robinhood fell from 19.6% to 0.0%, down 19.6 points across the baseline window. Pionex.US sits at 0.1%, down 0.6 points, having moved 0.1 point up from September on a single valid recommendation in 673 observations.

PayPal and Robinhood hold no valid recommendations in the October qualified set. Pionex.US holds one valid recommendation, matching its presence in the broader series only at trace levels. Neither PayPal nor Robinhood has recorded prior-to-current movement since their absence began, which the series shows as a flat reading rather than a further decline.

The distinction: absence is not the same as low ranking. A brand can appear in an answer without being recommended; PayPal and Robinhood are not appearing in the qualified set at all. For Pionex.US, the single valid recommendation is the entire basis for its 0.1% reading and should not be read as a recovery.

Highest-priority diagnostic: For Robinhood, which prompt clusters and surfaces stopped producing any presence, and is the gap a coverage issue or a change in what AI systems can find?

Buyer-Intent Interpretation

Questions This Section Answers

• Which buyer-intent clusters currently have qualified observations, and which are empty?
• What can the benchmark not currently answer about fee-based and head-to-head buyer questions?

In May 2026 the qualified observations split across three clusters: 1,069 in brand recommendation, 98 in pricing and value, and 424 in multi-brand comparison. From July 2026 onward, all qualified observations fell into the brand recommendation cluster, and the pricing and comparison clusters are empty in the October series.

The public benchmark therefore cannot currently answer the fee-based and head-to-head questions buyers actually ask. October's recommendation-shaped answer share of 51.7% is the strongest the series has recorded, but all of that strength sits in one cluster. The vertical has a clear picture of who gets recommended and no current picture of who wins when the prompt is about cost or a direct comparison. That matters commercially in a category where fee sensitivity and exchange switching are common buyer behaviours.

Brand Opportunity Summary

Questions This Section Answers

• What is each brand's highest-priority diagnostic based on its October coverage signal?
• Which brands need investigation into why they are shortlisted but rarely ranked first?

The benchmark identifies where attention is warranted; a company-level analysis is needed to explain why.

Evidence Behind the Benchmark

The aggregate metrics are built from prompt-level observations (query, surface, recommendation outcome, rank, sentiment, and citations where exposed). Company-level analysis can go deeper into prompt, competitor, surface, and evidence patterns. Source presence is not automatically treated as proof of causation.

About This Benchmark

This report is part of the CiteWorks Studio AI Market Discovery research program.

• AI Visibility Industry Market Methodology
• AI Visibility Industry Market Metrics
• AI Visibility Industry Market Standards

Report-Specific Interpretation Notes

• Small-count movement. Pionex.US moved from 11 valid recommendations in May 2026 to one in October 2026, and Uphold and Uniswap each carry fewer than 130 valid recommendations in the month; these absolute counts should anchor any reading of their percentage movement.
• Qualified denominator versus raw collection. October's 673 qualified observations come from 800 collected prompts across six AI surface families. The May baseline published 1,591 qualified observations without a recorded prompt funnel, so month-to-month percentage comparisons are made within the qualified set in each month.
• Directional analysis. Month-over-month movement identifies changes worth investigating; it does not by itself establish the cause of those changes. The August contraction and the September and October recovery describe the distribution of AI output, not the mechanism behind the shifts.
• Conflicting published claims. The inconsistency alerts describe platforms that gave incompatible answers to the same question. They identify a factual risk in AI output, not an established ranking or a settled fee schedule.

Next Step

The Public Benchmark Shows Where a Brand Is Winning or Losing. A Company-Level Audit Shows Why.

The aggregate percentages conceal the questions that decide commercial outcomes: which high-intent prompts a brand wins, which competitor takes the recommendation when it loses, what attributes AI systems associate with each option, and which external sources shape those answers. October's numbers say Kraken is recommended first more often than anyone else and that Binance, Crypto.com, and Gemini are almost always included but rarely first. They do not say which prompt, which surface, or which evidence source produced that pattern, and the five platform conflicts reported this month show how much variance sits underneath a single monthly figure.

A company-specific AI visibility audit maps those prompt, surface, competitor, ranking, sentiment, and evidence-source patterns into a prioritized visibility strategy. It answers why one brand converts presence into top placement while another does not, and where a competitor is being named instead.

Request an AI visibility audit