The State of Agent Visibility in Fintech 2026
How visible are the world's top fintech companies to AI agents?
Published by Slobodan Manić / Web Performance Tools
Measurement date: May 25, 2026
More than one in three top fintech sites is partially invisible to AI crawlers
We measured 274 companies from the CNBC World's Top Fintech Companies 2025 list (the most comprehensive fintech-specific ranking available) using the Agent Visibility Curve (AVC), a measurement framework that quantifies how much of a page's content is accessible to AI agents under real-world crawling conditions.
- 99 of 274 fintech sites (36%) deliver less than 80% of their homepage content in raw HTML, which is the view that GPTBot, ClaudeBot, and most production AI crawlers receive.
- 55 sites (20%) deliver less than 30% of their content without JavaScript.
- 47 sites (17%) deliver zero content to AI crawlers that don't execute JavaScript.
- Once rendered with a browser for 5 seconds, 273 of 274 sites (99%) deliver 80% or more of their content. The gap is entirely a JavaScript-execution problem.
- The median fintech site takes 21x longer to render than to fetch as raw HTML.
These findings are based on a single measurement per site taken on May 25, 2026, from Portugal via residential connection, using the WPT Agent Visibility Validator. This is a snapshot, not a trend.
Severity distribution across 274 sites
274 fintech sites measured
More than one in three top fintech sites is partially invisible to AI crawlers
99 of 274 measurable fintech sites (36%) deliver less than 80% of their homepage content in raw HTML. 47 sites (17%) deliver zero content to AI crawlers without JavaScript execution. Their homepages are functionally nonexistent to GPTBot, ClaudeBot, and similar crawlers.
| Severity | Count | Percentage |
|---|---|---|
80%+ visible | 175 | 64% |
60-79% visible | 24 | 9% |
30-59% visible | 20 | 7% |
Under 30% visible | 55 | 20% |
Raw visibility distribution
Rendering closes the gap for nearly all sites
273 of 274 sites (99%) deliver 80% or more of their content once rendered with a browser for 5 seconds. The content exists but is locked behind JavaScript execution. For the vast majority, 5 seconds of rendering is enough to surface everything.
Three sites do not reach 100% content visibility after 5 seconds of rendering. One remains below 80%, two are above 80% but still incomplete.
| Company | Raw HTML | At 5 seconds | At network idle |
|---|---|---|---|
| Niyo | 0% | 0% | 100% |
| Lianlian DigiTech | 0% | 90.5% | 100% |
| Au Jibun Bank | 88.0% | 92.5% | 100% |
All three are Asian companies (India, China, Japan) measured from Portugal. Niyo's content appears only after network idle at 6 seconds. Lianlian DigiTech and Au Jibun Bank have high raw fetch times (962ms, 1,741ms) that consume a large portion of the 5-second render budget on network latency alone. Measured from a closer geographic origin, these sites would likely complete within the window.
Rendering is expensive
Time to idle ratio measures how much longer a page takes to reach network idle (when all background network activity stops) compared to fetching the raw HTML. A ratio of 21x means the browser took 21 times longer to reach idle than the raw fetch took to complete.
The median fintech site has a time to idle ratio of 19.7x. AI systems choosing whether to render a page face a direct tradeoff between compute cost and content completeness.
Time to idle vs. raw fetch by percentile
34 fintech sites never reach network idle
34 of 274 sites (12%) do not reach network idle within 30 seconds. Persistent background activity (analytics, polling, live updates) keeps network connections open indefinitely. A rendering crawler that gives these pages 30 seconds still cannot determine when the page is "done."
Time from page load to network idle (no open connections for 500ms). Sites exceeding 30 seconds hit the measurement timeout.
47 fintech sites deliver zero content without JavaScript
All 47 zero-content sites return an HTML shell that JavaScript populates after load. An AI crawler fetching the raw HTML receives nothing.
The time to idle ratio shows how much longer the page takes to reach network idle compared to fetching the raw HTML. A ratio of 498.2x (Plus500) means reaching idle took 498 times longer than the raw fetch. Sorted by ratio, highest first.
| Company | Raw fetch | Time to idle ratio |
|---|---|---|
| Plus500 | 61ms | 498.2x |
| Fireblocks | 72ms | 420.4x |
| C2FO | 84ms | 363.3x |
| SymphonyAI | 74ms | 283.2x |
| Tipalti | 78ms | 133.6x |
| YouTrip | 61ms | 109.8x |
| Binance | 61ms | 101.8x |
| ACI Worldwide | 106ms | 97.0x |
| CoinTracker | 72ms | 94.1x |
| Zest AI | 63ms | 90.2x |
| Sapiens International | 68ms | 85.2x |
| Docupace | 70ms | 80.0x |
| Hi Marley | 67ms | 78.4x |
| Zafin | 79ms | 76.6x |
| Bondora | 76ms | 76.2x |
| Mylo | 68ms | 75.6x |
| Soldo | 75ms | 72.7x |
| Cleo AI | 86ms | 72.3x |
| SEON | 81ms | 71.6x |
| Coinbase | 80ms | 70.8x |
| Convera | 79ms | 70.8x |
| Justt | 80ms | 69.2x |
| C6 Bank | 75ms | 69.1x |
| LendingTree | 74ms | 68.8x |
| ICapital | 90ms | 63.0x |
| Moneybox | 87ms | 62.5x |
| Bkash | 87ms | 62.1x |
| Tyro | 85ms | 60.4x |
| DriveWealth | 91ms | 58.1x |
| ComplyAdvantage | 98ms | 54.6x |
| Capitolis | 111ms | 54.3x |
| TerraPay | 100ms | 51.5x |
| MoneySuperMarket | 126ms | 46.9x |
| Revolut | 180ms | 28.9x |
| Raize | 218ms | 24.1x |
| Toast | 264ms | 20.6x |
| Trading 212 | 262ms | 20.2x |
| Tazapay | 289ms | 18.7x |
| Wakam | 315ms | 16.9x |
| Lianlian DigiTech | 962ms | 16.0x |
| PalmPay | 362ms | 14.4x |
| Roadzen | 473ms | 11.6x |
| Ascend Money | 511ms | 10.4x |
| Navan | 553ms | 9.7x |
| OSL Group | 785ms | 6.9x |
| Niyo | 978ms | 6.2x |
| Ant Group | 1123ms | 5.4x |
Leaderboard: 100% visibility, fastest response
These 10 companies deliver 100% of their homepage content in raw HTML with the fastest raw fetch times in the study. An AI crawler gets everything on the first request, in under 121 milliseconds.
| Rank | Company | Raw visibility | Raw fetch time |
|---|---|---|---|
| 1 | Fiserv | 100.0% | 58ms |
| 2 | Acorns | 100.0% | 76ms |
| 3 | Trustly | 100.0% | 89ms |
| 4 | Ledger | 100.0% | 100ms |
| 5 | Kueski | 100.0% | 113ms |
| 6 | Galaxy | 100.0% | 114ms |
| 7 | Backbase | 100.0% | 131ms |
| 8 | Vitesse | 100.0% | 135ms |
| 9 | Rapyd | 100.0% | 137ms |
| 10 | Klar | 100.0% | 140ms |
Companies with 100% raw visibility
101 of 274 measurable fintech companies (37%) deliver all of their homepage content in raw HTML, with no JavaScript execution required. These sites are fully visible to every AI crawling strategy, from the simplest HTTP fetch to the most sophisticated rendering crawler.
Listed alphabetically. All scored equally.
Request your score
If your company is in the study, we'll send your individual results: raw visibility, rendered visibility, time to idle ratio, and how you compare to the median. If you're not on the list, we'll run a free Agent Visibility Check on your homepage and send you the results.
How we measured
The Agent Visibility Curve (AVC) measures how much of a page's content is accessible under three conditions:
1. Raw visibility (no JavaScript)
This is what GPTBot, ClaudeBot, and most production AI crawlers see. Text is extracted from <main>, <article>, or <body> as fallback (only using <main> or <article> if they actually contain text) and converted to Markdown, preserving structural elements that AI systems use when ingesting web content.
2. Rendered visibility (5 seconds)
Content visible after 5 seconds of browser rendering from TTFB. This approximates rendering crawlers with realistic time budgets. Content is extracted from the live DOM using the same Markdown conversion.
3. Rendered full (network idle)
Content captured after the page reaches network idle (no open connections for 500ms), or after a 30-second timeout. DOM mutations tracked as diagnostic only. This is the 100% denominator.
Severity thresholds
| Rating | Raw visibility |
|---|---|
Full visibility | 80% or above |
Partial visibility | 60% to 79% |
Low visibility | 30% to 59% |
Near-zero visibility | Below 30% |
Measurement protocol
Each homepage was measured once on May 25, 2026. The measurement consists of a raw HTTP fetch (no JavaScript) followed by a full browser render using Chromium via Playwright. The browser ran in non-headless mode to minimize bot protection interference.
Technical environment
AVC was developed as part of the Machine-First Architecture framework authored by Slobodan Manić.
Sample
Source: CNBC World's Top Fintech Companies 2025. Only the canonical homepage was measured per site.
Limitations
- Single geographic origin (Portugal). Sites may behave differently from other regions.
- Single measurement per site. Transient conditions may affect individual results.
- Homepage only. Interior pages may have different visibility characteristics.
- No scrolling or interaction. Content gated behind scroll or click is not captured.
The State of Agent Visibility in Fintech 2026
A research publication by Slobodan Manić, published through Web Performance Tools. The Agent Visibility Curve (AVC) methodology was developed as part of the Machine-First Architecture framework.
Based on measurements taken May 25, 2026, from Portugal via residential connection, using the WPT Agent Visibility Validator v0.1.0. Each site was measured once. Sites may perform differently under different conditions.
Download per-company measurements for all 300 companies (CSV).
Download CSV
Slobodan Manić
Co-founder, Web Performance Tools
Author of Machine-First Architecture, a framework for building websites that are visible to both human visitors and AI agents. He has 15 years of experience in web development and optimization, including a stint as CTO of Search Engine Journal, and is a five-time WordPress core contributor. He hosts the No Hacks podcast (220+ episodes) and speaks at conferences including Experimentation Elite, WordCamps, Marketing Festival, and Conversion Hotel.
Need help improving your score?
Web Performance Tools offers Agent Readiness Audits for enterprise teams. We measure exactly what AI agents see when they visit your site and deliver a prioritized roadmap your engineering team can execute.