Audible's best-seller chart is one of the few public signals of what people buy in audiobooks. Indie authors checking comp titles, narrators watching their genre, and "what's hot in audio" newsletters all want that chart in a script, with the ratings behind each title.
This post builds audible.py, one Python file that uses only the standard library (urllib, sqlite3, re). It does five things:
-
probecompares the chart page with the "best sellers" sort of Audible's catalog API (they don't match, and that's the first trap) -
chartreads a best-seller chart and joins it with rating data -
splitcompares narration scores with story scores across several genre charts -
search,book,reviews,author,categories, andbestsellerslook up the catalog through an API -
tracksaves chart snapshots in SQLite and prints rank moves and new ratings between runs
Everything below ran on November 16, 2026 on a Linux VM with Python 3.13.5, and every output block is pasted unedited from those runs.
Read this first. Audible's terms prohibit scraping. The Audible Service Conditions of Use (last updated November 2, 2026) incorporate the Amazon.com Conditions of Use and say Audible "is an Amazon Service for the purposes of" them. The "License and Access" section of those Amazon conditions (last updated August 14, 2026) says your license "does not include any resale or commercial use of any Amazon Service, or its contents; any collection and use of any product listings, descriptions, or prices; any derivative use of any Amazon Service or its contents; [...] or any use of data mining, robots, or similar data gathering and extraction tools." Audible's own "Content Restrictions" add that you may not "use the Service or Content for any commercial or illegal purpose." So collecting Audible data with a script goes against its terms, whoever runs the requests. Read the terms yourself and weigh the legal risk for your use case before you run anything here. If an official source covers what you need, use it instead (see the next section). Reviews are personal writing by real listeners, so don't republish them, and store only what you need.
Is there an Audible API?
Not a public, documented one for catalog data. There's no developer portal, no key signup, and no published terms for third-party catalog access. Here's what does exist:
- If you're the author, narrator, or rights holder, your real sales and royalty numbers are in your ACX or publisher dashboard. Those are actual sales, which the public chart never shows. If that's your use case, stop here.
-
Your own library. The Audible apps talk to
api.audible.com/1.0/.... Account endpoints need a registered device. The open-sourceaudiblePython package handles that registration for your own account, which is the reasonable way to export your own library. -
Catalog endpoints answer without a login.
api.audible.com/1.0/catalog/productsreturned product data, ratings, and reviews to a plain script with no key. It's undocumented, so it can change or close at any time, and the terms quoted above still apply.
Also check robots.txt before writing any loop. For www.audible.com it has a block labeled #Block excess charts with Disallow: /charts/*?. That means the query-less chart URLs (/charts/best and /charts/best/<category>/<id>) are allowed and anything with a query string, including ?page=2, is not. Review pages like /allprodreviews are disallowed too. The script below only fetches page one of each chart, which is the top 20.
Step 1: Setup
No pip install. The header defines a descriptive User-Agent, the endpoints, and a small get() that backs off on 429/5xx:
#!/usr/bin/env python3
"""Audible best-seller charts, ratings, reviews and a chart tracker. Python 3.10+, stdlib only."""
import html
import json
import os
import re
import sqlite3
import sys
import time
import urllib.error
import urllib.parse
import urllib.request
UA = "audible-chart-notes/1.0 (personal research; contact: you@example.com)"
WEB = "https://www.audible.com"
CATALOG = "https://api.audible.com/1.0/catalog/products"
SA_API = "https://api.scrapeatlas.com/v1/audible"
SA_KEY = os.environ.get("SCRAPEATLAS_API_KEY", "")
GROUPS = "product_desc,rating,product_attrs,contributors,series,category_ladders"
DB = os.environ.get("AUDIBLE_DB", "audible.db")
def get(url, params=None, headers=None, tries=4):
"""GET with a real User-Agent and backoff on 429/5xx. Returns (status, text)."""
if params:
url += "?" + urllib.parse.urlencode(params)
for attempt in range(tries):
req = urllib.request.Request(url, headers={"User-Agent": UA, **(headers or {})})
try:
with urllib.request.urlopen(req, timeout=120) as r:
return r.status, r.read().decode("utf-8", "replace")
except urllib.error.HTTPError as e:
body = e.read().decode("utf-8", "replace")
if e.code in (429, 500, 502, 503) and attempt < tries - 1:
wait = 2 ** attempt * 3
print(f" {e.code} on {url[:70]}... retry in {wait}s", file=sys.stderr)
time.sleep(wait)
continue
return e.code, body
raise RuntimeError("unreachable")
The User-Agent matters. When I sent an empty User-Agent, /charts/best answered 503 both times I tried. With the Python default (Python-urllib/3.x) or a descriptive string like the one above, it answered 200. Put a real contact in yours.
Step 2: Parse the chart page
https://www.audible.com/charts/best is server-rendered HTML, so you don't need a browser. Each title sits in an element with id="product-list-item-<ASIN>", the visible rank is in an <h3> ("1.", "2.", ...), and the title is in an aria-label attribute. Splitting the page on the item ID is simpler and sturdier than a full HTML parser here:
def chart_url(category=None):
# robots.txt has "Disallow: /charts/*?" -> only fetch query-less chart URLs (page 1, 20 titles)
return f"{WEB}/charts/best" + (f"/{category}" if category else "")
def parse_chart(page):
rows = []
for block in page.split('id="product-list-item-')[1:]:
asin = block[:10]
title = re.search(r"aria-label='([^']*)'", block)
rank = re.search(r'<h3[^>]*>\s*<span[^>]*>\s*(\d+)\.\s*</span>', block)
rows.append({
"rank": int(rank.group(1)) if rank else None,
"asin": asin,
"title": html.unescape(html.unescape(title.group(1))) if title else None, # double-escaped in markup
})
return rows
def fetch_chart(category=None):
status, page = get(chart_url(category))
if status != 200:
raise SystemExit(f"chart page returned HTTP {status}")
rows = parse_chart(page)
if not rows:
raise SystemExit("chart page parsed to 0 rows: markup changed, look at the HTML")
return rows
Two things I hit while writing this:
-
Titles are double-escaped. The
aria-labelcontainsCarl&#39;s Doomsday Scenario, so onehtml.unescape()leaves you withCarl's. You need to unescape twice. -
data-positionrestarts on every page. Each item also has adata-positionattribute. On a?page=2URL it starts again at 1, so it isn't the chart rank. (I fetched page 2 once while exploring, before I'd readrobots.txt. The script doesn't.) Use the visible<h3>rank.
fetch_chart() exits loudly if the page parses to zero rows. A silent empty chart in a tracker looks exactly like "nothing changed".
Step 3: The catalog API, and why its "best sellers" sort isn't the chart
The chart page doesn't give you exact rating counts, the narration/story split, release dates, or runtime. The catalog API does, and it accepts up to 20 ASINs per call via asins=. You choose the fields with response_groups. Without product_desc you get no title at all: my first run crashed with KeyError: 'title' until I added it.
def catalog(asins):
"""Batch lookup, 20 ASINs per call. Returns {asin: product}."""
out = {}
for i in range(0, len(asins), 20):
chunk = asins[i:i + 20]
status, body = get(CATALOG, {"asins": ",".join(chunk), "response_groups": GROUPS})
if status != 200:
raise SystemExit(f"catalog API HTTP {status}: {body[:200]}")
for p in json.loads(body).get("products", []):
out[p["asin"]] = p
time.sleep(1)
return out
def catalog_sorted(sort="BestSellers", n=20, category=None):
params = {"products_sort_by": sort, "num_results": n, "response_groups": GROUPS}
if category:
params["category_id"] = category
status, body = get(CATALOG, params)
if status != 200:
raise SystemExit(f"catalog API HTTP {status}: {body[:200]}")
return json.loads(body)
def dist(p, kind="overall"):
d = (p.get("rating") or {}).get(f"{kind}_distribution") or {}
return d.get("average_rating") or 0.0, d.get("num_ratings") or 0
The catalog API also has a products_sort_by=BestSellers option, and it's tempting to treat that as the chart. probe puts the two side by side:
def cmd_probe():
"""Is the API's BestSellers sort the same list as the chart page? (It isn't.)"""
chart = fetch_chart()
api = catalog_sorted("BestSellers", 20)
api_asins = [p["asin"] for p in api["products"]]
print(f"chart page: {len(chart)} titles | catalog sort=BestSellers: {len(api_asins)} of {api['total_results']:,}")
print(f"{'#':>2} {'chart page':<42} {'API BestSellers sort':<42}")
for i in range(10):
c = chart[i]["title"]
a = api["products"][i]["title"] + f" ({api['products'][i].get('release_date')})"
print(f"{i + 1:>2} {short(c, 42):<42} {short(a, 42):<42}")
overlap = set(api_asins) & {r["asin"] for r in chart}
print(f"overlap in top 20: {len(overlap)}")
$ python3 audible.py probe
chart page: 20 titles | catalog sort=BestSellers: 20 of 66,012
# chart page API BestSellers sort
1 Exoplanet Love Me Wicked: Zeke Wicked (The Wickeds:…
2 Dungeon Crawler Carl To Crown It All (2026-11-17)
3 A Court of Splintered Harmony Bro Doll (2027-02-16)
4 Daddy Issues Blood of the Bastion (2026-11-14)
5 Carl's Doomsday Scenario The Washwoman at the River's End (2027-03…
6 Mad Mabel Building the Marriage You Want (2027-01-2…
7 Sunrise on the Reaping (The Hunger Games) A Majestic Christmas (The Billionaire Bro…
8 The Dungeon Anarchist's Cookbook Last Ride (2026-12-08)
9 The Gate of the Feral Gods Deck the Halls with Boughs of Murder (202…
10 The Witch Dragon Bound 4: A Legacy Building Fantasy…
overlap in top 20: 0
Zero overlap in the top 20. The API's "BestSellers" sort is full of books released in the last few days or still on pre-order (release dates out to March 2027), most with 0 to 6 ratings. Whatever that sort ranks by, it isn't the public chart. If you write about "Audible best sellers", the chart page is the source that matches what readers see.
The total_results figure isn't stable either. For the same BestSellers query I saw 65,243 at about 10:59, 66,012 at 11:03, and 74,865 at 11:05 (UTC+7, all on November 16). Don't treat it as a catalog size.
Step 4: The chart with real numbers
chart takes the 20 ASINs from the page and looks them up in one batch call:
def cmd_chart(category=None):
rows = fetch_chart(category)
meta = catalog([r["asin"] for r in rows])
print(f"{chart_url(category)} ({len(rows)} titles)")
print(f"{'#':>2} {'title':<40} {'ratings':>8} {'overall':>7} {'perf':>5} {'story':>5} {'hrs':>5} released")
for r in rows:
p = meta.get(r["asin"], {})
o, n = dist(p)
perf, _ = dist(p, "performance")
story, _ = dist(p, "story")
hrs = (p.get("runtime_length_min") or 0) / 60
print(f"{r['rank']:>2} {short(r['title'], 40):<40} {n:>8,} {o:>7.2f} {perf:>5.2f} {story:>5.2f} "
f"{hrs:>5.1f} {p.get('release_date', '?')}")
missing = [r["asin"] for r in rows if r["asin"] not in meta]
if missing:
print("not returned by catalog API:", ", ".join(missing))
$ python3 audible.py chart
https://www.audible.com/charts/best (20 titles)
# title ratings overall perf story hrs released
1 Exoplanet 5,686 4.44 4.70 4.34 4.1 2026-11-05
2 Dungeon Crawler Carl 73,987 4.84 4.92 4.80 13.5 2021-01-28
3 A Court of Splintered Harmony 6,108 4.38 4.69 4.29 12.0 2026-10-27
4 Daddy Issues 1,463 4.88 4.82 4.88 6.3 2026-11-10
5 Carl's Doomsday Scenario 42,193 4.89 4.96 4.86 11.5 2021-04-22
6 Mad Mabel 20,920 4.79 4.86 4.77 9.3 2026-04-21
7 Sunrise on the Reaping (The Hunger Game… 39,401 4.85 4.82 4.83 12.8 2025-03-18
8 The Dungeon Anarchist's Cookbook 40,430 4.89 4.96 4.85 16.9 2021-05-14
9 The Gate of the Feral Gods 33,302 4.90 4.97 4.88 18.1 2021-09-16
10 The Witch 9,764 4.49 4.57 4.46 9.5 2026-10-06
11 The Butcher's Masquerade 35,809 4.94 4.98 4.92 23.6 2022-05-26
12 The Eye of the Bedlam Bride 25,796 4.91 4.97 4.89 26.8 2023-09-01
13 This Inevitable Ruin 31,933 4.93 4.97 4.90 28.7 2025-02-11
14 Theo of Golden 39,731 4.84 4.88 4.81 13.2 2025-12-02
15 A Parade of Horribles 33,382 4.92 4.98 4.89 20.4 2026-05-12
16 The Demon-Haunted World 5,812 4.67 4.69 4.67 17.4 2017-05-30
17 Department of Fate 28 4.89 4.93 4.96 15.8 2026-11-10
18 Contend 1,670 4.97 4.96 4.92 7.8 2026-10-13
19 Bad in Bed 11,013 4.78 4.88 4.71 8.7 2026-10-22
20 Threshing Day (Wing and Claw Collection) 9,737 4.79 4.84 4.76 5.9 2026-09-29
A few things stand out:
- Rank isn't popularity. Department of Fate is #17 with 28 ratings, released six days earlier. Dungeon Crawler Carl, from January 2021, is #2 with 73,987. The chart reflects recent sales, while rating counts pile up over years, so don't use one as a stand-in for the other.
- One author takes 8 of the 20 spots. Matt Dinniman's Dungeon Crawler Carl books are at #2, #5, #8, #9, #11, #12, #13, and #15. Series backlists move together on Audible, and a chart tracker should group by series or author, not only by title.
-
There are three averages, not one. Every title has
overall,performance, andstorydistributions. That's the most Audible-specific thing in the data, and Step 5 uses it.
Step 5: Narration vs. story
On Audible, listeners rate the performance (the narrator) separately from the story. That lets you ask a question you can't ask about print books: where does the narrator lift the book, and where do they hold it back? split reads several genre charts, removes repeats (a title can appear on the overall chart and two genre charts), skips titles with fewer than 200 ratings on either axis, and sorts by performance - story:
def cmd_split(categories):
"""Narration vs story: where do listeners rate the performance above the book?"""
rows, seen, dupes = [], set(), 0
for cat in categories:
chart = fetch_chart(cat or None)
meta = catalog([r["asin"] for r in chart])
for r in chart:
p = meta.get(r["asin"])
if r["asin"] in seen: # same title on several charts: count it once
dupes += 1
continue
seen.add(r["asin"])
if not p:
continue
perf, pn = dist(p, "performance")
story, sn = dist(p, "story")
if min(pn, sn) < 200: # tiny samples swing wildly
continue
narrators = ", ".join(n["name"] for n in p.get("narrators", [])[:2])
label = cat.split("/")[0].removesuffix("-audiobooks") if cat else "overall"
rows.append((perf - story, perf, story, min(pn, sn), r["title"], narrators, label))
time.sleep(2)
rows.sort(reverse=True)
print(f"{len(seen)} unique titles on {len(categories)} charts ({dupes} repeats skipped); "
f"{len(rows)} have >=200 performance and story ratings\n")
print(f"{'gap':>6} {'perf':>5} {'story':>5} {'n':>8} title / narrator")
for row in rows[:8] + [None] + rows[-5:]:
if row is None:
print(" ...")
continue
g, perf, story, n, title, narr, cat = row
print(f"{g:>+6.2f} {perf:>5.2f} {story:>5.2f} {n:>8,} {short(title, 36)} / {short(narr, 28)} [{short(cat, 20)}]")
gaps = [r[0] for r in rows]
print(f"\nperformance > story for {sum(g > 0 for g in gaps)} of {len(gaps)} titles; "
f"median gap {sorted(gaps)[len(gaps) // 2]:+.3f}")
The category paths come from the links on the chart page itself (/charts/best/<slug>/<id>). I used the overall chart plus six genres:
$ python3 audible.py split "" science-fiction-fantasy-audiobooks/18580606011 \
mystery-thriller-suspense-audiobooks/18574597011 romance-audiobooks/18580518011 \
biographies-memoirs-audiobooks/18571951011 business-careers-audiobooks/18572029011 \
history-audiobooks/18573518011
106 unique titles on 7 charts (34 repeats skipped); 91 have >=200 performance and story ratings
gap perf story n title / narrator
+0.59 4.58 3.99 415 All She Wants for Christmas / Ruth Wilson, Matthew Rhys [mystery-thriller-su…]
+0.47 4.45 3.98 20,679 Yesteryear: A GMA Book Club Pick / Rebecca Lowman [mystery-thriller-su…]
+0.44 4.34 3.91 274 Caller Unknown / Juliet Rylance [mystery-thriller-su…]
+0.40 4.69 4.29 5,980 A Court of Splintered Harmony / Elizabeth Evans [overall]
+0.39 4.82 4.43 284 Split Second: Thirty-Two Switcheroo / Lorelei King [mystery-thriller-su…]
+0.37 4.70 4.34 5,605 Exoplanet / Anthony Mackie, Ariana DeBo… [overall]
+0.30 4.72 4.42 631 Christmas Staycation / Kelsey Navarro Foster, Bran… [romance]
+0.29 4.51 4.22 961 The Single Father / Tamaryn Payne [mystery-thriller-su…]
...
-0.06 4.82 4.88 1,459 Daddy Issues / Sofia Franklyn [overall]
-0.06 4.76 4.82 35,060 Bad Blood / Will Damron [business-careers]
-0.09 4.57 4.66 2,190 Influence, New and Expanded / Robert B. Cialdini [business-careers]
-0.10 4.61 4.71 370 The Nerd Reich / Gil Durán [business-careers]
-0.14 4.55 4.69 53,310 Fourth Wing / Rebecca Soler, Teddy Hamilt… [science-fiction-fan…]
performance > story for 73 of 91 titles; median gap +0.082
From this one snapshot:
- Narrators usually score higher. Performance beat story for 73 of 91 titles, but the median gap is small (+0.082 stars).
- Thrillers top the list. Four of the five biggest positive gaps are on the mystery/thriller chart. A story average just under 4.0 with performance above 4.3 is a book that listeners rate for its narrator.
- Business and memoir go the other way. Business books and an author-read memoir (Daddy Issues) are among the few where story beats performance. Fourth Wing has the largest negative gap of the sample (-0.14) on 53,310 ratings, so that's not small-sample noise.
- Keep the 200-rating floor. Without it, Department of Fate (28 ratings) would show up with a confident-looking gap from a handful of people.
This is one day's chart, not a study. To say anything about narrators in general, run it across many snapshots and categories.
Step 6: Searches, single titles, reviews, and authors through an API
Past the chart, the edge cases pile up: ASINs differ by marketplace, reviews come from a separate endpoint whose raw payload includes the reviewer's Amazon account ID (which you shouldn't store), and author pages and the category tree are more undocumented shapes that can change without notice.
So the rest of the script uses ScrapeAtlas*, a scraping API that covers 50 platforms, Audible among them. It returns the same catalog data as normalized JSON (rating.overall/performance/story, authors[], categories[] ladders), takes marketplace as a parameter (us, uk, ca, au, in, de, fr, it, es, jp), and returns review objects without the account ID. The helper sends an x-api-key header and turns error responses into readable messages:
def sa(path, **params):
if not SA_KEY:
raise SystemExit("Set SCRAPEATLAS_API_KEY first.")
status, body = get(f"{SA_API}/{path}", params, headers={"x-api-key": SA_KEY})
data = json.loads(body)
if status != 200:
failure = data.get("failure") or {}
raise SystemExit(f"ScrapeAtlas {path} HTTP {status}: {failure.get('code')} {failure.get('message')}")
return data
def short(s, n):
s = s or ""
return s if len(s) <= n else s[: n - 1] + "…"
I tested six Audible routes: search, audiobook, reviews, author, categories, and best-sellers.
Search
def cmd_sa_search(query):
data = sa("search", query=query, sort="relevance", limit=5)
print(f"search {query!r}: {data['total']} results (showing {len(data['results'])})")
for b in data["results"]:
o = b["rating"]["overall"]
print(f" {b['asin']} {short(b['title'], 44):<44} {o['count']:>8,} ratings avg {o['average']:.2f} "
f"{b['runtimeMinutes'] / 60:.1f}h {b['releaseDate']}")
$ python3 audible.py search dungeon crawler carl
search 'dungeon crawler carl': 154 results (showing 5)
B08V8B2CGV Dungeon Crawler Carl 73,987 ratings avg 4.84 13.5h 2021-01-28
B0934GTSGT Carl's Doomsday Scenario 42,193 ratings avg 4.89 11.5h 2021-04-22
B094XCNV6G The Dungeon Anarchist's Cookbook 40,430 ratings avg 4.89 16.9h 2021-05-14
B09GCYRZRQ The Gate of the Feral Gods 33,302 ratings avg 4.90 18.1h 2021-09-16
B0GL147D7S Cello's Gate 28 ratings avg 4.71 18.9h 2026-11-03
The fifth result, Cello's Gate, is a different book released 2026-11-03, with 28 ratings. Title search matches on words, so check the author before you trust a top-N list.
One title with all three distributions
def cmd_sa_book(asin, marketplace="us"):
b = sa("audiobook", asin=asin, marketplace=marketplace)["audiobook"]
print(f"{b['title']} [{b['asin']}, {marketplace}]")
print(f" by {', '.join(a['name'] for a in b['authors'])} | read by {', '.join(n['name'] for n in b['narrators'])}")
print(f" {b['publisher']} | released {b['releaseDate']} | {b['runtimeMinutes'] // 60}h{b['runtimeMinutes'] % 60:02d}m | {b['formatType']}")
for kind in ("overall", "performance", "story"):
d = b["rating"][kind]
bars = " ".join(f"{s}*:{d[w]:,}" for s, w in zip("54321", ("five", "four", "three", "two", "one")))
print(f" {kind:<11} avg {d['average']:.3f} from {d['count']:,} ({bars})")
print(f" written reviews: {b['rating']['reviewCount']:,}")
print(" categories:", " | ".join(" > ".join(c["name"] for c in ladder) for ladder in b["categories"]))
$ python3 audible.py book B0GP1KB3T5
Daddy Issues [B0GP1KB3T5, us]
by Sofia Franklyn | read by Sofia Franklyn
Simon & Schuster Audio | released 2026-11-10 | 6h19m | unabridged
overall avg 4.879 from 1,463 (5*:1,345 4*:81 3*:21 2*:10 1*:6)
performance avg 4.818 from 1,459 (5*:1,292 4*:101 3*:40 2*:19 1*:7)
story avg 4.875 from 1,459 (5*:1,341 4*:78 3*:22 2*:12 1*:6)
written reviews: 252
categories: Biographies & Memoirs > Entertainment & Celebrities | Biographies & Memoirs > Women
Only 252 written reviews against 1,463 ratings. Most Audible listeners give stars without writing anything, so a review sample tells you about the people who write, not about everyone who rated.
ASINs are per marketplace
$ python3 audible.py book B08V8B2CGV de; python3 audible.py book B08V893CH7 de
ScrapeAtlas audiobook HTTP 404: not_found The Audible audiobook, author or category was not found.
--- same book, audible.de's English-edition ASIN:
Dungeon Crawler Carl [B08V893CH7, de]
by Matt Dinniman | read by Jeff Hays
Audible Studios | released 2021-01-28 | 13h31m | unabridged
overall avg 4.775 from 1,802 (5*:1,530 4*:187 3*:49 2*:23 1*:13)
performance avg 4.910 from 1,472 (5*:1,380 4*:68 3*:13 2*:6 1*:5)
story avg 4.702 from 1,472 (5*:1,185 4*:190 3*:63 2*:14 1*:20)
written reviews: 159
categories: Literatur & Belletristik > Belletristik > Mit Videospiel-, Film- & Fernsehbezug | Krimis & Thriller > Thriller | Science Fiction & Fantasy > Fantasy > Paranormal & Urban Fantasy > Zeitgenössische Fantasy
The US ASIN for Dungeon Crawler Carl doesn't exist on audible.de: the English edition sold there is B08V893CH7, and there's a German translation (B0G4RYRYB3) as well. ScrapeAtlas gives a clean 404 not_found for the wrong one. The raw catalog API is less helpful: api.audible.de/1.0/catalog/products/B08V8B2CGV answered HTTP 200 with {"product":{"asin":"B08V8B2CGV"},"response_groups":["always-returned"]}, which is an empty product, not an error. If you go direct, check for a title before you trust a 200. Also notice the German listeners' split for the same audiobook: performance 4.910 vs. story 4.702.
Reviews: recent vs. critical vs. everyone
def cmd_sa_reviews(asin, pages=3):
"""Pull 'critical' and 'recent' reviews and compare the sample with the full distribution."""
for sort in ("recent", "critical"):
reviews, meta = [], None
for page in range(1, pages + 1):
data = sa("reviews", asin=asin, sort=sort, page=page, limit=20)
meta = meta or data
reviews += data["reviews"]
if not data["hasMore"]:
break
time.sleep(1)
stars = [r["overallRating"] for r in reviews if r["overallRating"]]
low = sum(s <= 2 for s in stars)
print(f"sort={sort}: {len(reviews)} reviews, {low} rated 1-2 stars "
f"({low / max(len(stars), 1):.0%}), dates {min(r['submittedAt'] for r in reviews)[:10]}"
f" .. {max(r['submittedAt'] for r in reviews)[:10]}")
for r in reviews[:3]:
print(f" {r['overallRating']}/{r['performanceRating']}/{r['storyRating']} "
f"{short(r['title'].strip(), 40)!r} helpful={r['helpfulVotes']}")
d = meta["rating"]["overall"]
print(f"all ratings: {(d['one'] + d['two']) / d['count']:.1%} are 1-2 stars "
f"({d['one'] + d['two']:,} of {d['count']:,})")
$ python3 audible.py reviews B0HDFGKSTL
sort=recent: 60 reviews, 14 rated 1-2 stars (23%), dates 2026-11-15 .. 2026-11-16
5/5/5 'Time is an insane !' helpful=0
5/5/5 'WOW' helpful=0
4/5/4 'Good listen' helpful=0
sort=critical: 60 reviews, 60 rated 1-2 stars (100%), dates 2026-11-09 .. 2026-11-15
1/3/1 "that's some pretty bad writing" helpful=0
1/2/1 "One of the worst books I've ever read" helpful=0
1/2/1 'Why would Andy Weir produce this?' helpful=0
all ratings: 6.8% are 1-2 stars (385 of 5,687)
This is the most useful output in the post if you analyze review sentiment. Exoplanet was the #1 title on the chart that morning:
- All 60 of the most recent reviews were posted within about a day (November 15 to 16). For a hot new release, "recent" means the last few hours, not a representative sample.
- 23% of those recent reviews are 1–2 stars, but only 6.8% of all 5,687 ratings are. People who write reviews in launch week are much more negative than the star ratings overall. If you build a "sentiment" metric from the newest reviews, it'll read as much worse than the book's actual rating.
-
sort=criticalis exactly what it says: all 60 reviews were 1–2 stars. It's useful for "what do people complain about", useless as a sample. - Notice the scores like
1/3/1: a critical reviewer can still rate the narration higher than the story. That's the Step 5 split again, one listener at a time.
Authors and categories
def cmd_sa_author(asin):
data = sa("author", asin=asin, limit=50)
a = data["author"]
print(f"{a['name']} ({a['asin']}): {data['total']} titles on Audible US")
rows = []
for b in data["books"]:
o = b["rating"]["overall"]
rows.append((o["count"], b))
for n, b in sorted(rows, key=lambda x: -x[0])[:10]:
lang = "" if b["language"] == "english" else f" [{b['language']}]"
print(f" {n:>8,} {b['rating']['overall']['average']:.2f} {short(b['title'], 44)}{lang} {b['releaseDate']}")
def cmd_sa_categories():
data = sa("categories")
cats = data["categories"]
print(f"{len(cats)} top-level categories, {sum(len(c.get('children', [])) for c in cats)} children")
for c in cats[:6]:
print(f" {c['id']} {c['name']} ({len(c.get('children', []))} children)")
$ python3 audible.py author B002BLP1QY
Matt Dinniman (B002BLP1QY): 38 titles on Audible US
73,987 4.84 Dungeon Crawler Carl 2021-01-28
42,193 4.89 Carl's Doomsday Scenario 2021-04-22
40,430 4.89 The Dungeon Anarchist's Cookbook 2021-05-14
35,809 4.94 The Butcher's Masquerade 2022-05-26
33,382 4.92 A Parade of Horribles 2026-05-12
33,302 4.90 The Gate of the Feral Gods 2021-09-16
31,933 4.93 This Inevitable Ruin 2025-02-11
25,796 4.91 The Eye of the Bedlam Bride 2023-09-01
7,172 4.59 Operation Bounce House 2026-02-10
2,705 4.35 Dominion of Blades 2017-06-22
$ python3 audible.py categories
24 top-level categories, 240 children
18571910011 Arts & Entertainment (6 children)
18571951011 Biographies & Memoirs (15 children)
18572029011 Business & Careers (6 children)
18572091011 Children's Audiobooks (20 children)
24427740011 Comedy & Humor (3 children)
18573211011 Computers & Technology (6 children)
The author list explains Step 4: the eight Dungeon Crawler Carl audiobooks on the chart are the author's eight most-rated titles. The category IDs are the same ones in the chart URLs, so you can build /charts/best/<slug>/<id> for any of the 24 top-level categories without scraping the navigation.
The best-sellers route has the same catch as Step 3
def cmd_sa_bestsellers(category=None, marketplace="us"):
params = {"limit": 10, "marketplace": marketplace}
if category:
params["categoryId"] = category
data = sa("best-sellers", **params)
print(f"best-sellers {marketplace} category={category}: total {data['total']:,}")
for i, b in enumerate(data["results"], 1):
print(f" {i:>2}. {short(b['title'], 46):<46} {b['rating']['overall']['count']:>7,} ratings released {b['releaseDate']}")
$ python3 audible.py bestsellers
best-sellers us category=None: total 74,865
1. Love Me Wicked: Zeke Wicked (The Wickeds: Dar… 2 ratings released 2026-11-13
2. To Crown It All 0 ratings released 2026-11-17
3. Bro Doll 0 ratings released 2027-02-16
4. Blood of the Bastion 0 ratings released 2026-11-14
5. The Washwoman at the River's End 0 ratings released 2027-03-23
6. Building the Marriage You Want 0 ratings released 2027-01-26
7. A Majestic Christmas (The Billionaire Brother… 6 ratings released 2026-11-13
8. Last Ride 0 ratings released 2026-12-08
9. Deck the Halls with Boughs of Murder 0 ratings released 2026-11-24
10. Dragon Bound 4: A Legacy Building Fantasy 5 ratings released 2026-11-13
This is the same order as the catalog API's BestSellers sort from Step 3 (Love Me Wicked, To Crown It All, Bro Doll, ...), not the chart page. To be clear about what I found: it's useful for spotting new and upcoming releases that sell early. If you need the public top-20 chart, read the chart page as in Step 2 and use the API only to add data to it. That's what the tracker does.
Step 7: Track the chart in SQLite
A single snapshot only tells you the order right now. The tracker stores each chart's 20 rows with rating and review counts, then compares with the previous snapshot for that chart. It prints entries, exits, rank moves, and how many ratings and reviews each title gained:
SCHEMA = """
CREATE TABLE IF NOT EXISTS snap (taken TEXT, chart TEXT, rank INTEGER, asin TEXT, title TEXT,
ratings INTEGER, avg REAL, reviews INTEGER,
PRIMARY KEY (taken, chart, asin));
"""
def cmd_track(categories):
db = sqlite3.connect(DB)
db.executescript(SCHEMA)
taken = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())
for cat in categories:
name = cat or "overall"
prev_at = db.execute("SELECT max(taken) FROM snap WHERE chart=?", (name,)).fetchone()[0]
prev = {r[0]: r[1:] for r in db.execute(
"SELECT asin, rank, ratings, reviews, title FROM snap WHERE chart=? AND taken=?", (name, prev_at))}
rows = fetch_chart(cat or None)
meta = catalog([r["asin"] for r in rows])
print(f"[{name}] {taken} vs {prev_at or 'nothing (first run)'}")
for r in rows:
p = meta.get(r["asin"], {})
avg, n = dist(p)
reviews = (p.get("rating") or {}).get("num_reviews") or 0
db.execute("INSERT INTO snap VALUES (?,?,?,?,?,?,?,?)",
(taken, name, r["rank"], r["asin"], r["title"], n, avg, reviews))
if not prev:
continue
if r["asin"] not in prev:
print(f" NEW #{r['rank']:<2} {short(r['title'], 44)} ({n:,} ratings)")
continue
old_rank, old_n, old_rev, _ = prev[r["asin"]]
if old_rank != r["rank"] or n != old_n:
move = old_rank - r["rank"]
arrow = f"up {move}" if move > 0 else f"down {-move}" if move < 0 else "same"
print(f" #{r['rank']:<2} ({arrow:>7}) {short(r['title'], 40):<40} ratings {n - old_n:+,} reviews {reviews - old_rev:+,}")
for asin, (old_rank, _, _, title) in prev.items():
if asin not in {r["asin"] for r in rows}:
print(f" OUT was #{old_rank:<2} {short(title, 44)}")
if not prev:
print(f" stored {len(rows)} rows")
db.commit()
time.sleep(2)
First run (overall chart plus Science Fiction & Fantasy):
$ python3 audible.py track "" science-fiction-fantasy-audiobooks/18580606011
[overall] 2026-11-16T04:06:04Z vs nothing (first run)
stored 20 rows
[science-fiction-fantasy-audiobooks/18580606011] 2026-11-16T04:06:04Z vs nothing (first run)
stored 20 rows
Second run, 23 minutes later:
$ python3 audible.py track "" science-fiction-fantasy-audiobooks/18580606011
[overall] 2026-11-16T04:29:01Z vs 2026-11-16T04:06:04Z
#1 ( same) Exoplanet ratings +6 reviews +1
#4 ( same) Daddy Issues ratings +1 reviews +0
#6 ( same) Mad Mabel ratings +3 reviews +1
#7 ( same) Sunrise on the Reaping (The Hunger Game… ratings +6 reviews +0
#9 ( same) The Gate of the Feral Gods ratings +1 reviews +0
#10 ( same) The Witch ratings +4 reviews +1
#14 ( same) Theo of Golden ratings +2 reviews +0
#15 ( same) A Parade of Horribles ratings +1 reviews +0
#18 ( same) Contend ratings +2 reviews +0
#19 ( same) Bad in Bed ratings +3 reviews +0
#20 ( same) Threshing Day (Wing and Claw Collection) ratings +1 reviews +0
[science-fiction-fantasy-audiobooks/18580606011] 2026-11-16T04:29:01Z vs 2026-11-16T04:06:04Z
#1 ( same) Exoplanet ratings +6 reviews +1
#6 ( same) The Gate of the Feral Gods ratings +1 reviews +0
#10 ( same) A Parade of Horribles ratings +1 reviews +0
#11 ( same) Threshing Day (Wing and Claw Collection) ratings +1 reviews +0
#12 ( same) Project Hail Mary ratings +2 reviews +0
Not a single rank changed in 23 minutes, on either chart, so the chart doesn't update in real time. Ratings did move: Exoplanet gained 6 ratings and 1 written review in that window (and shows up on both charts, since it's #1 in each), Sunrise on the Reaping gained 6, and titles that gained nothing are left out of the diff. Over a day, the "ratings gained" column is the more interesting one: it's a rough proxy for listening activity, which the chart's rank alone doesn't give you.
To run it daily, add a cron entry (or a systemd timer) and keep the cadence low. The charts don't need minute-level polling, and robots.txt already limits you to page one:
15 7 * * * cd /home/you/audible && SCRAPEATLAS_API_KEY=... python3 audible.py track "" science-fiction-fantasy-audiobooks/18580606011 >> track.log 2>&1
The tracker only needs the chart page and the catalog API; the key matters only if you add Step 6 commands to the job. After a few weeks, query days on chart per title, best rank per series, and ratings gained per day vs. rank.
The CLI
if __name__ == "__main__":
cmd, *args = sys.argv[1:] or ["help"]
cats = args or [""]
commands = {
"probe": lambda: cmd_probe(),
"chart": lambda: cmd_chart(args[0] if args else None),
"split": lambda: cmd_split(cats),
"search": lambda: cmd_sa_search(" ".join(args)),
"book": lambda: cmd_sa_book(*args),
"reviews": lambda: cmd_sa_reviews(args[0]),
"author": lambda: cmd_sa_author(args[0]),
"bestsellers": lambda: cmd_sa_bestsellers(*args),
"categories": lambda: cmd_sa_categories(),
"track": lambda: cmd_track(cats),
}
if cmd not in commands:
raise SystemExit("usage: audible.py probe|chart [cat]|split [cat...]|search Q|book ASIN [mkt]|"
"reviews ASIN|author ASIN|bestsellers [catId] [mkt]|categories|track [cat...]")
commands[cmd]()
Gotchas I hit, collected
- The terms prohibit scraping. The Audible conditions bring in Amazon's "no data mining, robots" clause. Read them before running anything.
-
robots.txt allows
/charts/...without a query string only. That's page one, the top 20 per category. -
"BestSellers" sort ≠ the chart. It showed zero overlap with the chart's top 20, and its
total_resultsjumped between 65,243 and 74,865 in six minutes. -
No
product_desc, no title.response_groupsdecides which fields you get back. -
An empty User-Agent got
503on the chart page both times I tried it. -
Titles are double-escaped in
aria-label, anddata-positionrestarts on every page. -
ASINs are per marketplace. The raw API answers a wrong-marketplace ASIN with an empty
200. - Recent reviews aren't a sample. For new releases they cluster in the last day and run far more negative than the ratings.
-
Raw review payloads include
author_id. Drop it if you collect reviews at all. - Rank ≠ rating count. A six-day-old book with 28 ratings sat at #17.
Wrapping up
- The public chart is HTML. Read the top 20 per category, and let the catalog API fill in exact ratings, runtime, and release dates.
- Audible's three rating axes (overall, performance, story) are its most distinctive data. In this snapshot narration beat story for 73 of 91 titles, and thrillers showed the biggest gaps.
- Treat reviews with suspicion. Writers are a minority of raters, and launch-week reviewers are harsher.
- Two SQLite snapshots show rank moves and new ratings, which is the closest public proxy for sales momentum.
What would you track on Audible: your own genre's chart, narrators, or new releases? And if you've found a better public signal for audiobook sales than the chart, I'd like to hear it.
*I build it.
Top comments (0)