Railway’s “AI recommends us” playbook, applied to CurrencyTransfer
14 Oct 2026 · Sources: Ahrefs Site Explorer & Keywords Explorer (UK, 13 Oct 2026), live crawl of currencytransfer.com. Estimates are in orange italics. Builds on Growth Analyst’s “transfer money to spain” AI Overview pilot (9–15 Oct). Nothing on the site was changed.
Verdict: no to “thousands more how-to pages”. CT already runs Railway’s tactic #3: 2,702 programmatic pages (812 currency-pair + 1,890 corridor pages). The corridor set brings in about 1,221 visits a month in total, roughly 0.6 per page. The currency set brings 166.5k a month (75% of site traffic), but it ranks for lookups like “pesos to dollars” and “dollar to rand”, not for anyone moving £10k+ from the UK or Gulf. Instead: prune and deepen what exists, build about 60 high-intent pages (property abroad, comparisons/alternatives, bank fees, Gulf→UK), and make CT machine-readable for AI agents (llms-full.txt, a quote endpoint, an MCP/ChatGPT app). AI recommendations reward being the most citable, specific source, not having the most pages.
Where CT stands today
| Metric | Value | Source |
|---|
| Organic traffic / keywords | 222,310 visits a month · 29,341 keywords (4,467 in top 3) | Ahrefs, subdomains |
| /currencies/ (812 URLs) | 166,525 visits a month · 21,661 keywords. Top pages: MXN→USD 7.0k, USD→ZAR 6.7k, EUR→ILS 3.8k, TRY→USD 3.6k | Ahrefs prefix |
| /countries/ corridors (1,890 URLs) | 1,221 visits a month · 349 keywords · 15 in top 3 | Ahrefs prefix |
| llms.txt | ✅ live (7 KB, good ICP positioning: £10k+, Gulf→UK). ❌ no llms-full.txt, ❌ no .md page versions | crawl |
| robots.txt | ✅ allows GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot | crawl |
| Structured data | 3 JSON-LD blocks on homepage; corridor/FAQ schema not audited | crawl |
| Comparison / alternative pages | ❌ none found (/compare/, /wise-alternative/ return 404). sitemap-banks.xml is empty | crawl |
| AI share of voice vs Wise, Revolut, OFX etc. | Not measured. The Ahrefs account has no Brand Radar add-on (“Missing addon: Brand Radar”), so there is no baseline yet. The Growth Analyst pilot is the only live AI-visibility test. | Ahrefs API error |
Evidence from competitors: Wise’s 15 biggest pages are all localised currency-converter pages (e.g. usd-to-idr 858k visits a month, jpy-to-vnd 642k). Rate pages work at Wise scale because Wise has the brand, local languages and a real converter behind them. CT’s pages win the same long-tail lookups but get none of the conversions.
Railway tactic → CT equivalent
| # | Railway | CT equivalent | Status |
|---|
| 1–2 | Get LLMs to recommend by default; GTM owns it | Own the prompts “best way to send £50k+ abroad”, “buy property in Spain from UK/Dubai currency”, “Wise alternative for large transfers”. Track them weekly | Pilot running (Spain) |
| 3 | 6,653 how-to pages | Already 2,702 pages. Fix quality rather than add more | Over-built, under-useful |
| 4 | 127 framework guides | ~25 use-case guides: property in ES/PT/FR/UAE/CY, salary repatriation Gulf→UK, supplier payments, inheritance, school fees | Gap |
| 5 | Comparison + migration vs every rival | “CT vs Wise/Revolut/OFX/TorFX/Currencies Direct/Moneycorp/Xe/WorldFirst/Airwallex” plus “[X] alternative for large transfers” | Gap (0 pages) |
| 6 | llms.txt + plain-text pages | Add llms-full.txt and .md twins of the ~100 pages that matter | Half done |
| 7 | Agent-oriented docs | Public, cached “indicative quote” JSON plus a docs page an agent can follow | Gap |
| 8 | Plugins in ChatGPT/Claude | MCP server / ChatGPT app: get_indicative_rate, compare_providers, request_callback | Gap |
| 9 | Zero-signup one command | Instant indicative quote without signup, then hand off to the regulated partner | Check current flow |
| 10 | Weekly changelog | Weekly “GBP/AED/EUR transfer outlook” + monthly “bank fees index” updates (freshness signal) | Blog posts ~22/mo already; repoint them |
Page-set sizing (UK, Ahrefs)
| Set | Example keywords (UK volume / KD) | Pages | Addressable traffic | Call |
|---|
| GBP currency pairs | gbp to usd 484k/32 · gbp to eur 67k/34 · gbp to aud 37k/1 | exists (812) | Huge, but intent is checking a rate. Traffic potential 34k (USD), 59k (EUR) | Keep; add a “sending £10k+? compare providers” module |
| UK→country corridors | india 2.5k/8 · pakistan 2.3k/6 · nigeria 1.2k/4 · australia 900/1 · france 800/0 · usa 800/3 · spain (transfer) 600/6 · poland 350/1 · portugal 250 · dubai 200 | exists (1,890) | ~10k/mo across the top 10. Remittance corridors (IN/PK/NG) are not ICP | Prune to ~40 ICP corridors, deepen them, noindex the rest (est.) |
| Property abroad | buying property in spain from uk 150/2 (traffic potential 20,000) | ~15 new | low volume per query, but the highest-value ICP clicks | Build: #1 priority |
| Comparison / alternatives | wise vs revolut 800/0 · wise alternative 250/0 (CPC $12) · torfx review 150 · moneycorp review 100 · currencies direct review 70 | ~20 new | ~1.5k/mo direct; often cited by AI | Build |
| Bank fee pages | barclays intl transfer fees 100 (TP 1,100) · hsbc 20 (TP 7,700) · natwest 10 (TP 300) | ~12 new (fills empty sitemap-banks) | ~9k traffic potential across the cluster | Build with real, dated fee tables |
| Generic head terms | international money transfer 2.2k/40 · best way to send money abroad 450 (TP 27k) · currency broker 250 (TP 3.9k) · business intl payments 200 (CPC $20) | ~5 hubs | Competitive; matters most for AI citations | Strengthen existing hubs |
| B2B how-to | pay overseas supplier 10 · ofx vs wise 10 | — | Negligible search volume in the UK | Skip as a programmatic set |
Rough total from net-new high-intent sets: ~60 pages, a few thousand visits a month, but at CPCs of $2–20 versus ~$0.35 for rate lookups. The value comes from conversions and AI citations, not volume.
Ranked plays
| # | Play | Effort | Impact |
|---|
| 1 | Property-abroad + Gulf→UK use-case guides (ES, PT, FR, CY, IT, UAE; salary repatriation) with worked £-examples, partner coverage, and timelines. Extends the Spain pilot | M | High |
| 2 | Comparison/alternative set vs 10 rivals: honest tables (min transfer, rate margin, dealer access, forward contracts) | M | High (AI cites these) |
| 3 | Prune/consolidate the 1,890 corridors to ~40 ICP routes; add a large-transfer comparison block to the top 50 currency pages | M | High (quality signal + conversions) |
| 4 | Agent layer: llms-full.txt, .md twins, public indicative-quote JSON, then an MCP server / ChatGPT app | M–L | Medium now, High later |
| 5 | Bank-fee index (populate sitemap-banks): dated fee tables for 12 UK + 6 UAE banks, refreshed monthly | S–M | Medium |
| 6 | Buy Brand Radar (or a similar tool) and track ~50 prompts weekly. Without it you can’t measure progress | S | Enabler |
Agent / MCP angle: keeping it introducer-safe
Offer read-only tools: get_indicative_rate(pair, amount), compare_providers(corridor, amount), request_introduction(contact, need). Don’t let the agent execute payments, take funds, or give personalised advice. Execution, KYC and the binding quote stay with the regulated partner (Moneycorp, Currencycloud etc.) under their own authorisation. Label rates as indicative and name the partner. Before launch, get compliance/legal sign-off that the agent flow is still an introduction and not arranging or a payment service (FCA PSRs 2017 / financial promotions). This is not legal advice.
Risks
- Scaled content abuse (Google spam policy): 1,890 near-identical corridor pages with about 0.6 visits each look like exactly what it targets. Pruning lowers the risk.
- YMYL finance: needs named, credentialed authors, dated fee data, sources, and regulatory disclosures on each page.
- Making pages genuinely useful: live rates, a worked example at £50k, a real fee comparison against a bank, which partner covers the route, and the documents needed.
- Off-ICP traffic inflates the numbers. Report pipeline from AI and organic, not sessions.
90-day plan
| Window | Do |
|---|
| Days 1–14 | Read the Spain pilot result (after 15 Oct). Set up prompt tracking. Audit the corridors: keep, merge or noindex. Ship llms-full.txt and .md twins for top pages |
| Days 15–45 | 6 property guides + 10 comparison pages. Large-transfer block on the top 50 currency pages |
| Days 46–75 | Bank-fee index. Weekly outlook cadence. Pilot an indicative-quote JSON endpoint with partner sign-off |
| Days 76–90 | MCP server / ChatGPT app MVP (read-only). Re-measure AI share of voice and pipeline. Decide on scaling |
First 3 actions
- Approve a Brand Radar (or equivalent) baseline across ~50 ICP prompts × ChatGPT/Perplexity/AI Overviews, so progress can be measured.
- Brief the 6 property-abroad guides, using the Spain pilot as the template.
- Export the 1,890 corridor URLs with Ahrefs traffic and mark each keep/merge/noindex. Target: ~40 deep ICP corridors.