CrustAPI vs SerpApi benchmark
Run: July 11, 2026, from a single machine (macOS, residential US connection).
Method: identical queries sent to both APIs back to back with SerpApi's no_cache=true
(SerpApi serves cached results for repeated queries; caching would fake the timings),
wall-clock time measured around each HTTPS request. Keys redacted. Script at the bottom.

== WEB SEARCH (4 unique queries, run once each) ==
Queries: "coffee roasters brooklyn", "how to start an llc",
         "running shoes for flat feet", "b2b lead generation tools"

serpapi  GET serpapi.com/search.json?engine=google&no_cache=true
  timings (ms): 4698, 815, 2937, 1422   -> median 2180
  organic results per query: 9, 6, 6, 8

crustapi GET crustapi.com/v1/search?type=web
  timings (ms): 1245, 1065, 1028, 1036  -> median 1051
  organic results per query: 8, 7, 7, 8

Note the variance: SerpApi swung from 0.8s to 4.7s across four queries.
CrustAPI stayed within 1.0 to 1.2s.

== IMAGES ("latte art", 1 query each, no_cache) ==
serpapi  8282 ms, 100 images returned
crustapi 1657 ms, 99 images returned (from the July 11 serper benchmark run)

== QUOTA / BILLING RECEIPT ==
SerpApi account API (serpapi.com/account.json), same key, before and after the run:
  before: this_month_usage: 0, total_searches_left: 250
  after:  this_month_usage: 6, total_searches_left: 244
Every search counted against the monthly quota, including the probe that Google
fuzzy-matched. The account object also carries plan_renewal_date: the quota resets
monthly. SerpApi is a subscription: Free 250/mo, Starter $25 for 1,000/mo,
Developer $75 for 5,000/mo, Production $150 for 15,000/mo, Big Data $275 for
30,000/mo (their public pricing page, July 11, 2026). Their FAQ describes unused
searches converting to credits only as a plan-downgrade mechanism.

CrustAPI is prepaid credits: purchased credits never expire, there is no renewal
date, and a zero-result or failed search is not billed.

== HONEST NOTES ==
- One machine, one day, four queries. SerpApi's spread (0.8s to 4.7s) means single
  runs vary a lot; run the script yourself before deciding anything.
- SerpApi returned an ai_overview field on some queries, which CrustAPI does not
  have. They also cover 80+ engines beyond Google and offer their Legal US Shield.
- Our zero-result billing probe was inconclusive for SerpApi: Google fuzzy-matched
  the nonsense phrase and returned 9 results, so that search was legitimately
  successful. We make no claim about SerpApi's zero-result billing.

== SCRIPT (python3, stdlib only; put your own keys in) ==

import json, time, urllib.request, statistics
from urllib.parse import quote

SERPAPI_KEY='...'; CRUST_KEY='...'
QUERIES=['coffee roasters brooklyn','how to start an llc',
         'running shoes for flat feet','b2b lead generation tools']

def serpapi(q):
    t0=time.time()
    d=json.loads(urllib.request.urlopen(
        f'https://serpapi.com/search.json?engine=google&no_cache=true&q={quote(q)}&api_key={SERPAPI_KEY}',
        timeout=60).read())
    return (time.time()-t0)*1000, d

def crust(q):
    req=urllib.request.Request(
        f'https://crustapi.com/v1/search?type=web&q={quote(q)}',
        headers={'x-api-key': CRUST_KEY})
    t0=time.time()
    d=json.loads(urllib.request.urlopen(req, timeout=60).read())
    return (time.time()-t0)*1000, d

s,c=[],[]
for q in QUERIES:
    ms,_=serpapi(q); s.append(ms)
    ms,_=crust(q); c.append(ms)
print('serpapi median', statistics.median(s))
print('crust   median', statistics.median(c))