Chain of Thought: Podcast Growth Audit

Executive Summary

The public content foundation is strong. The five latest episode pages contain substantial transcript text, chapters, and listening links. Preserve that work.
The clearest fix is transcript name accuracy. The latest Slack and Neat conversations contain guest-name variants that conflict with the page headings. Correct those before adding more discovery features.
Test clearer packaging and subscription destinations. The Cisco title can state the security problem more directly. All five sampled pages mix the show and host YouTube handles; clarify the intended destination.
Listen Notes adds external context. Its accessible listings show COT at Listen Score 34 versus 45–50 for the selected peers. These are estimated-popularity signals from differently aged snapshots, not download counts; the linked comparison below explains the limits.

Pilot goal: grow a relevant audience and support sponsorship. This is an assumed goal for this demonstration. The scorecard describes observed readiness; it does not predict audience growth or revenue. Sources: public website and the episode links in the findings below.

Transcript pages / 5 sampled

Of the five sampled episode pages. Presence only; audio completeness was not checked.

Of the five sampled episode pages. Presence only; audio completeness was not checked.Source: Five reviewed episode pagesTable: reviewed_pages

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT COUNT(*) AS pages, SUM(transcript) AS transcripts, SUM(chapters) AS chapters, SUM(name_issue) AS flagged FROM reviewed_pages

Introductions flagged

Confirmed guest-name inconsistencies in recent introductions. This is not a full accuracy rate.

Confirmed guest-name inconsistencies in recent introductions. This is not a full accuracy rate.Source: Five reviewed episode pagesTable: reviewed_pages

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT COUNT(*) AS pages, SUM(transcript) AS transcripts, SUM(chapters) AS chapters, SUM(name_issue) AS flagged FROM reviewed_pages

Releases in 183 days

COT releases in the 183 days ending September 4, 2026.

COT releases in the 183 days ending September 4, 2026.Source: Release counts recomputed from saved RSS publication datesTable: rss_episodes

Counts published RSS entries in the 183-day window. rss_episodes is loaded from the saved public RSS snapshot by build_report.py.

SELECT show, COUNT(*) AS releases FROM rss_episodes WHERE julianday(published_at) >= julianday('2026-03-05T23:11:31+00:00') AND julianday(published_at) <= julianday('2026-09-04T23:11:31Z') GROUP BY show ORDER BY show

Pilot readiness assessment

This pilot uses evidence statuses instead of an uncalibrated total out of 100. “Strong” means the sampled evidence supports the criterion; “Fix” identifies a verified inconsistency; “Test” is an editorial hypothesis; “Context needed” requires the owner's intended goal or schedule. A missing source never becomes a failing score.

Listen Notes gives a directional popularity comparison

COT's accessible listing shows a Listen Score of 34 and a Top 3% global band. The peer listings show 45–50. These are Listen Notes' proprietary popularity signals, not our grade and not audience or download counts. The global band compares podcasts across all genres, regions and languages; this peer set is only a topical reference.

Scores were retrieved September 4 from public listing snapshots with different freshness. COT's opened listing was cached with an August 12 latest episode, so do not treat these values as a synchronized current ranking. Listen Notes says scores update about monthly; its methodology page reports a September 1 update. That does not establish when each cached listing's score was refreshed. How Listen Score works.

Source: Public Listen Notes listing observationsTable: listen_notes_scores

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT show, score, global_band, listing_url FROM listen_notes_scores ORDER BY score DESC, show

Publicly visible Listen Scores

Source: Public Listen Notes listing observationsTable: listen_notes_scores

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT show, score, global_band, listing_url FROM listen_notes_scores ORDER BY score DESC, show

Publicly visible Listen Scores data
Show	Listen Score (0–100)	Global band
Practical AI	50	TOP 0.5%
No Priors	47	TOP 1%
Latent Space	45	TOP 1%
Chain of Thought	34	TOP 3%

Explore the peer listings

Practical AI: Listen Score 50, Top 0.5% global band.
No Priors: Listen Score 47, Top 1% global band.
Latent Space: Listen Score 45, Top 1% global band.
Chain of Thought: Listen Score 34, Top 3% global band.

No public listener rating/count was displayed in the accessible listings. Use the live RSS snapshot for publication dates and catalog counts: the COT and Practical AI listing extracts differed from their feeds. Do not translate score gaps into listener ratios.

Source: Public Listen Notes listing observationsTable: listen_notes_scores

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT show, score, global_band, listing_url FROM listen_notes_scores ORDER BY score DESC, show

Correct guest names before expanding the transcript library

Availability is already good; editorial accuracy needs a separate check. The Neat page identifies the guest as Tormod Ree, but the transcript introduction renders his name as “Tormid Reed” and later “Torment.” The Slack page identifies Jaime DeLanghe, while its transcript introduction says “Jamie DeLange.” These are text inconsistencies verified against the same pages, not an assessment of the audio.

Review names, organizations, product terms and numbers against the recording before publishing a transcript. Start with these two introductions and propagate confirmed corrections to the website and any separately hosted transcript. Check the surrounding segment rather than globally replacing ordinary words.

Sources: Neat conversation and Slack conversation.

Source: Chain of Thought public website, September 4, 2026File: evidence/web.json

Reviewed homepage, start-here, subscribe, robots, sitemap index and five latest episode HTML pages. Text and structured data collected without browser interaction.

Test a title that states the Cisco episode's security problem

The current title is “Parenting Your AI Agents for Best Results | Cisco's Jeetu Patel.” The notes and chapters contain a more specific promise: dynamic runtime guardrails, agent trust, and testing proposed fixes before production.

Suggested test: “Why Static Security Rules Fail AI Agents | Cisco's Jeetu Patel.” This is an editorial hypothesis grounded in the episode, not a proven improvement. Keep the underlying claim attributed to the guest and verify it against the recording before publishing a replacement.

Test the framing in comparable promotional placements. If impression and click data are available, compare click-through rates within the same channel and comparable audiences. A simple before/after change to a published title cannot isolate the title's effect.

Source: Cisco conversation.

Source: Chain of Thought public website, September 4, 2026File: evidence/web.json

Reviewed homepage, start-here, subscribe, robots, sitemap index and five latest episode HTML pages. Text and structured data collected without browser interaction.

Make the show subscription destination consistent

The paths exist; the labels need a decision. Every sampled episode page links both @ChainOfThoughtAI and @ConorBronsdon on YouTube. The latter appears in the host connection block, so this may be intentional. It is not evidence of a broken link or lost subscriptions.

Use the show channel for a show-subscription call to action. If the host channel is also useful, label it explicitly as the host's channel. Verify whether the handles resolve to distinct channels before making a change.

The homepage already offers a newsletter with a stated benefit and frequency, and the subscribe page lists listening platforms. Preserve those paths. Owner-side click and signup data would be needed to diagnose actual conversion loss.

Sources: Subscribe and the five sampled episode pages.

Source: Chain of Thought public website, September 4, 2026File: evidence/web.json

Reviewed homepage, start-here, subscribe, robots, sitemap index and five latest episode HTML pages. Text and structured data collected without browser interaction.

Compare publishing patterns without treating volume as quality

The comparison set contains three AI shows serving overlapping topics. It is a convenience sample, not a matched cohort by age, budget, audience size, or format. Cadence means episodes published in the trailing 183 days divided by 183/30. Transcript coverage below means RSS items carrying a transcript tag; it says nothing about uninspected peer websites.

COT's release volume is close to Practical AI and No Priors in this window. More frequent publishing alone is not a reason to change the schedule. COT's high transcript-tag coverage is a distribution strength, while transcript accuracy remains a separate issue.

Source: Live RSS and Apple lookup snapshot, September 4, 2026File: evidence/benchmark.json

Public podcast-benchmark snapshot for four AI podcasts; cadence uses 183 days, transcript percentage counts RSS tags only.

Releases in the last 183 days

Source: Release counts recomputed from saved RSS publication datesTable: rss_episodes

Counts published RSS entries in the 183-day window. rss_episodes is loaded from the saved public RSS snapshot by build_report.py.

SELECT show, COUNT(*) AS releases FROM rss_episodes WHERE julianday(published_at) >= julianday('2026-03-05T23:11:31+00:00') AND julianday(published_at) <= julianday('2026-09-04T23:11:31Z') GROUP BY show ORDER BY show

Releases in the last 183 days data
Show	Releases
Chain of Thought	21
Latent Space	41
No Priors	24
Practical AI	23

Transcript tags make the feed easier to use

COT declares transcripts on 71 of 72 feed items (98.6%). Practical AI declares 66 of 372 (17.7%); the two other feeds checked do not expose this tag. This chart measures feed metadata only. It does not establish whether a peer publishes transcripts on its website, and it does not measure accuracy.

Source: Live RSS and Apple lookup snapshot, September 4, 2026File: evidence/benchmark.json

Public podcast-benchmark snapshot for four AI podcasts; cadence uses 183 days, transcript percentage counts RSS tags only.

RSS transcript-tag coverage

Source: Transcript tag coverage in the public RSS feedsTable: rss_episodes

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT show, COUNT(*) AS items, SUM(transcript_tag) AS tagged, AVG(transcript_tag) AS coverage FROM rss_episodes GROUP BY show ORDER BY coverage DESC, show

RSS transcript-tag coverage data
Show	Share of feed items	Tagged items	All feed items
Chain of Thought	98.6%	71	72
Practical AI	17.7%	66	372
Latent Space	0%	0	221
No Priors	0%	0	177

| Show | Releases / 183 days | Releases / 30 days | Feed items | RSS transcript tags (%) | |---|---:|---:|---:|---:| | Chain of Thought | 21 | 3.44 | 72 | 98.6 | | Latent Space | 41 | 6.72 | 221 | 0.0 | | Practical AI | 23 | 3.77 | 372 | 17.7 | | No Priors | 24 | 3.93 | 177 | 0.0 |

Source: Live RSS and Apple lookup snapshot, September 4, 2026File: evidence/benchmark.json

Public podcast-benchmark snapshot for four AI podcasts; cadence uses 183 days, transcript percentage counts RSS tags only.

Interpret release gaps in season context

COT published 21 episodes in the 183-day window. The five latest releases span August 12 through September 2; their calendar-day gaps are 7, 5, 3 and 6 days. The preceding July 16–August 12 gap spans 27 calendar days. A planned season break could explain it. Confirm the intended schedule before labeling it inconsistent or recommending more episodes.

The feed has 72 items, including one trailer, while the site presents 71 episodes. Keep feed-item counts distinct from editorial episode numbering. This audit does not treat the difference as a defect.

Days since the preceding release

Source: Calendar days between COT releasesTable: rss_episodes

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

WITH recent AS (SELECT date(published_at) AS published FROM rss_episodes WHERE show='Chain of Thought' ORDER BY published_at DESC LIMIT 10), gaps AS (SELECT published, CAST(julianday(published)-julianday(LAG(published) OVER (ORDER BY published)) AS INTEGER) AS days FROM recent) SELECT published, days FROM gaps WHERE days IS NOT NULL ORDER BY published

Days since the preceding release data
Release date (2026)	Calendar days
Jun 18	7
Jun 25	7
Jul 01	6
Jul 16	15
Aug 12	27
Aug 19	7
Aug 24	5
Aug 27	3
Sep 02	6

The publisher reports a cross-channel audience

The homepage displays the following counters in the September 4 snapshot. These are publisher-reported figures, not independently verified platform analytics. Downloads plus views combine different events and are not unique people. YouTube and newsletter subscribers may overlap; do not sum these cards. The newsletter figure is a stated lower bound. These provide context, not a growth score.

Source: Chain of Thought public website, September 4, 2026File: evidence/web.json

Reviewed homepage, start-here, subscribe, robots, sitemap index and five latest episode HTML pages. Text and structured data collected without browser interaction.

Downloads + views (reported)

158,825

Homepage total. Mixed events, not unique audience; reporting window not independently verified.

Homepage total. Mixed events, not unique audience; reporting window not independently verified.Source: Publisher-reported homepage countersTable: publisher_metrics

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT metric, value, qualifier FROM publisher_metrics ORDER BY metric

YouTube subscribers (reported)

3,020

Publisher-reported YouTube subscriber count as displayed September 4.

Publisher-reported YouTube subscriber count as displayed September 4.Source: Publisher-reported homepage countersTable: publisher_metrics

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT metric, value, qualifier FROM publisher_metrics ORDER BY metric

Newsletter subscribers (at least)

1,100

Homepage states 1,100+. A lower bound, not an exact count.

Homepage states 1,100+. A lower bound, not an exact count.Source: Publisher-reported homepage countersTable: publisher_metrics

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SELECT metric, value, qualifier FROM publisher_metrics ORDER BY metric

OP3 has no dataset for this show at the audit date

OP3 can supply measured downloads for shows tracked through its prefix. Downloads are not listeners, retention, or business leads, and tracking cannot recover downloads that occurred before the prefix was added.

Both the feed-URL and podcast-GUID lookups returned HTTP 404. The current feed contained 72 media enclosures and none had an OP3 prefix. No download totals, app shares or comparable episode windows were available from OP3. This is missing evidence, not zero downloads, and it does not lower the readiness assessment.

For a subsequent analytics review, use owner-authorized host analytics, or enable OP3 for future measurement after the owner chooses to do so. Compare completed seven-day download windows only when tracking covered each episode from release. Keep promotion, guest reach and release timing as possible explanations rather than attributing a difference to titles.

Source: OP3 lookup evidenceFile: evidence/op3.json

Bring your existing chapters into podcast apps

The transcript foundation is even stronger when the trailer is separated: all 71 full episodes declare a transcript. The one untagged feed item is the welcome trailer. These are presence checks; linked files and audio accuracy need separate validation.

| Feature | Observed coverage | Useful next step | |---|---:|---| | RSS transcript tags | 71 / 72 feed items; 71 / 71 full episodes | Preserve coverage and check transcript accuracy. | | Chapters on sampled website pages | 5 / 5 pages | Reuse the existing timestamps. | | RSS Podcasting 2.0 chapter tags | 0 / 72 items | Pilot external chapter JSON for the five sampled episodes. | | Episode-level person tags | 71 / 72 items | Check host and guest credits for the latest Slack conversation. |

Podcasting 2.0 chapters let supporting apps present navigation, links and optional chapter images. The external file can be corrected without replacing the audio. Check the host's publishing support, file accessibility and timestamps before rollout. Website timestamps alone do not establish app-native chapter support.

The latest Slack conversation is the sole item without episode-level person tags. Person metadata can identify hosts and guests; episode-level entries replace channel-level people, so include the host when adding a guest.

After these checks, consider a short preview using soundbite metadata and a curated podroll of related shows. Funding and payment features should follow the show's business model. Optional tag absence is an opportunity to assess, not an automatic failure.

Source: Independent Podcasting 2.0 RSS tag check, September 4, 2026File: evidence/podcasting2.json

Your episode pages have the foundations for discovery

The five sampled pages have titles, descriptions, matching canonical URLs, a single heading, episode schema and transcript sections. The score below measures those six specific checks on each page. It establishes a useful starting point for deeper search and AI visibility measurement.

Source: Six page-foundation checks on five recent episode pagesTable: page_checks

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(passed) AS passed, COUNT(*) AS assessed, 100.0 * SUM(passed) / COUNT(*) AS score FROM page_checks

Page foundations / 100

100

30 of 30 selected checks passed across five pages. Equal weighting; presence and structure only. This is not a ranking or AI visibility score.

30 of 30 selected checks passed across five pages. Equal weighting; presence and structure only. This is not a ranking or AI visibility score.Source: Six page-foundation checks on five recent episode pagesTable: page_checks

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(passed) AS passed, COUNT(*) AS assessed, 100.0 * SUM(passed) / COUNT(*) AS score FROM page_checks

| Check | Pages passing | What this establishes | |---|---:|---| | HTML title present | 5 / 5 | A page title is supplied | | Meta description present | 5 / 5 | A description is supplied | | One matching canonical | 5 / 5 | One declared canonical matches the fetched URL | | One nonempty H1 | 5 / 5 | One primary heading is present | | PodcastEpisode schema present | 5 / 5 | Parseable JSON-LD declares an episode | | Transcript section present | 5 / 5 | The HTML contains a transcript section |

Next: measure whether these pages earn impressions and clicks for relevant topics. A passing presence check does not assess title quality, transcript accuracy, schema completeness, indexing, crawl permissions, page speed or actual recommendations. The existing title experiment and name corrections remain useful despite this strong foundation score.

Source: Six page-foundation checks on five recent episode pagesTable: page_checks

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(passed) AS passed, COUNT(*) AS assessed, 100.0 * SUM(passed) / COUNT(*) AS score FROM page_checks

The show surfaces in search; its episode pages need a closer look

In a fixed eight-query panel, the owned website appeared in 6/8 searches (75%). All six appearances were the homepage. Neither of the two exact-title searches returned the owned episode page in the inspected window, although podcast-platform copies surfaced.

We inspected 12 returned web results per query using the search service available to this audit on September 4. This is a small observed panel, not a Google/Bing ranking report, traffic estimate or market-wide visibility score.

Source: Fixed eight-query public search panel and inspected link examplesTable: query_results

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(owned) AS owned_hits, AVG(owned) AS owned_rate, SUM(CASE WHEN exact_title=1 THEN episode ELSE 0 END) AS episode_hits FROM query_results

Website presence in 8-query panel

75%

6 of 8 queries returned an owned-domain URL among 12 inspected web results each. Fixed pilot sample; not search market share.

6 of 8 queries returned an owned-domain URL among 12 inspected web results each. Fixed pilot sample; not search market share.Source: Fixed eight-query public search panel and inspected link examplesTable: query_results

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(owned) AS owned_hits, AVG(owned) AS owned_rate, SUM(CASE WHEN exact_title=1 THEN episode ELSE 0 END) AS episode_hits FROM query_results

Owned episode-page hits / 2 title queries

Both exact-title queries returned podcast-platform copies; neither returned the owned episode page in the inspected window.

Both exact-title queries returned podcast-platform copies; neither returned the owned episode page in the inspected window.Source: Fixed eight-query public search panel and inspected link examplesTable: query_results

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(owned) AS owned_hits, AVG(owned) AS owned_rate, SUM(CASE WHEN exact_title=1 THEN episode ELSE 0 END) AS episode_hits FROM query_results

| Query | Owned website found | Interpretation | |---|---|---| | "Chain of Thought" podcast Conor Bronsdon | Homepage | Branded | | AI infrastructure podcast | Homepage | Category discovery | | AI agents podcast | Not in inspected results | Category discovery | | AI evals podcast | Homepage | Category discovery | | Jeetu Patel AI agents podcast | Not in inspected results | Guest and topic | | Jaime DeLanghe Slack podcast | Homepage | Guest and topic | | "Parenting Your AI Agents for Best Results" | Homepage | Exact episode title | | "Slack Wants to Be the Context Harness for Code" | Homepage | Exact episode title |

Next: inspect the two episode URLs in Search Console, verify their indexing and canonical selection, and strengthen relevant internal links to them. Include the canonical episode page in guest promotion materials. These are diagnostic next steps; absence from this result sample does not prove deindexing or explain its cause. Repeat the same queries to assess stability.

Source: Fixed eight-query public search panel and inspected link examplesTable: query_results

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(owned) AS owned_hits, AVG(owned) AS owned_rate, SUM(CASE WHEN exact_title=1 THEN episode ELSE 0 END) AS episode_hits FROM query_results

Track search exposure, AI citations and links separately

| Measurement | Useful numbers | Evidence needed | |---|---|---| | Search performance | Impressions, clicks, CTR, position, branded versus topic queries | Owner-authorized Search Console export with dates, country and device | | AI answer visibility | Answers mentioning the show; answers citing its site; prompt coverage | A fixed, unbranded prompt set, recorded answers, model/product versions and repeat runs | | First-party AI exposure | Cited pages and impressions/citations where the platform reports them | Google Search Console or Bing Webmaster Tools export | | Link authority | Referring domains, linking pages, linked episodes, newly observed/lost links | A dated backlink provider or Search Console Links export |

For the AI test, keep brand-recognition questions separate from discovery questions such as “Which podcasts explain production AI evaluation?” An answer that mentions a show and an answer that links to its website are different successes. Calculate each rate over completed test runs and show failures or unavailable runs separately. No such answer-rate score has been measured in this pilot.

For backlinks, distinguish independent editorial links from podcast-directory syndication and profiles controlled by the publisher. A public mention search can find useful examples, but cannot establish the site's total backlinks or domain authority. Provider scores should retain their provider name and date; they are not Google metrics.

Practical next step: establish a 28-day search and AI exposure baseline, identify episodes with impressions but few clicks, then pair those pages with the title/description tests. Track referring domains and links to individual episodes alongside it. This makes the next audit a progress report.

Measurement references: Google's AI Search insights announcement, Bing AI Performance, and Search Console Links report. These are available measurement paths; this report has no connected owner analytics or full backlink-index dataset.

Source: Fixed eight-query public search panel and inspected link examplesTable: query_results

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(owned) AS owned_hits, AVG(owned) AS owned_rate, SUM(CASE WHEN exact_title=1 THEN episode ELSE 0 END) AS episode_hits FROM query_results

Turn guest coverage into useful episode links

Five deliberately selected public pages reveal different kinds of links and mentions. They are examples to act on, not a complete backlink profile or a referring-domain score.

| Inspected source | Relationship | Observed destination | |---|---|---| | Conor Bronsdon podcast page | Host-owned site | Show website | | podcast-benchmark GitHub README | Host-owned repository | Show website | | Chain of Thought LinkedIn company profile | Show-controlled profile | Show website /linkedin-show route | | Scaled Cognition: Dan Klein episode article | Guest's company coverage | Apple and YouTube | | Galileo: Measuring AI ROI and Achieving Efficiency Gains | Former publisher coverage | pod.link and YouTube |

Useful opportunity: give each guest a short promotion kit containing the canonical episode URL, a description and an approved image. When a company writes about the interview, that URL offers readers the transcript, chapters and subscription choices in one place. Existing platform links can remain useful; a direct episode link is an additional destination to offer.

Measure total referring domains, new/lost links and provider authority scores only after importing a dated backlink dataset. Search Console and Bing exports or a verified-site Ahrefs account are possible next sources. No backlink total or authority score has been measured here.

Source: Fixed eight-query public search panel and inspected link examplesTable: query_results

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT SUM(owned) AS owned_hits, AVG(owned) AS owned_rate, SUM(CASE WHEN exact_title=1 THEN episode ELSE 0 END) AS episode_hits FROM query_results

Public reviews highlight accessible explanations

Apple's U.S. storefront displays a 5.0/5 average from 36 ratings. Four visible written reviews give an initial view of what listeners appreciate. Ratings and written reviews are separate measures; this is one storefront's public snapshot.

Source: Public listener feedback, by platform and storefrontTables: apple_feedback, review_coding

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT rating, ratings, sample FROM apple_feedback;
SELECT theme, COUNT(DISTINCT review_id) AS reviews FROM review_coding GROUP BY theme ORDER BY theme

Apple U.S. rating / 5

Public U.S. storefront average, based on 36 ratings.

Public U.S. storefront average, based on 36 ratings.Source: Public listener feedback, by platform and storefrontTables: apple_feedback, review_coding

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT rating, ratings, sample FROM apple_feedback;
SELECT theme, COUNT(DISTINCT review_id) AS reviews FROM review_coding GROUP BY theme ORDER BY theme

Apple U.S. ratings

Star ratings; this is not a written-review count.

Star ratings; this is not a written-review count.Source: Public listener feedback, by platform and storefrontTables: apple_feedback, review_coding

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT rating, ratings, sample FROM apple_feedback;
SELECT theme, COUNT(DISTINCT review_id) AS reviews FROM review_coding GROUP BY theme ORDER BY theme

Written reviews sampled

Four displayed U.S. review cards; the total written-review count is unavailable.

Four displayed U.S. review cards; the total written-review count is unavailable.Source: Public listener feedback, by platform and storefrontTables: apple_feedback, review_coding

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT rating, ratings, sample FROM apple_feedback;
SELECT theme, COUNT(DISTINCT review_id) AS reviews FROM review_coding GROUP BY theme ORDER BY theme

Themes in four displayed Apple U.S. reviews

Source: Public listener feedback, by platform and storefrontTables: apple_feedback, review_coding

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT rating, ratings, sample FROM apple_feedback;
SELECT theme, COUNT(DISTINCT review_id) AS reviews FROM review_coding GROUP BY theme ORDER BY theme

Themes in four displayed Apple U.S. reviews data
Manually coded theme	Reviews mentioning theme (sample: 4)
Accessible, informative explanations	2
Guest quality and practical AI use	2
Strong listening engagement or habit	3
Timely and relevant coverage	2

Themes overlap: one review can support several. The four displayed reviews are favorable and self-selected; their independence and representativeness are unverified. This is a small qualitative sample, not an audience sentiment score.

Use this: preserve accessible explanations and practical guest conversations in the show's positioning. Invite specific feedback about the episode that made someone return and the topic they want next. Link to the Apple U.S. listing for the original reviews.

Goodpods and Podchaser did not expose usable review totals in the accessible cached pages. Public YouTube fetches did not yield comment text. Those sources remain unavailable in this feedback analysis; no zero-review conclusion follows.

Source: Public listener feedback, by platform and storefrontTables: apple_feedback, review_coding

Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.

SELECT rating, ratings, sample FROM apple_feedback;
SELECT theme, COUNT(DISTINCT review_id) AS reviews FROM review_coding GROUP BY theme ORDER BY theme

Two live directory catalogs reconcile with the feed

The feed contains 71 full episodes plus one trailer. A later direct fetch confirms the September 2 release on both Apple and Listen Notes. The older Listen Notes search-cache observation in the popularity section remains the provenance for that earlier score snapshot.

| Directory | Catalog shown | Latest episode | Evidence status | |---|---|---|---| | Apple Podcasts | 71 full + trailer; lookup 72 | Sep 2, 2026 | Live page verified | | Listen Notes | 72 items; trailer semantics unspecified | Sep 2, 2026 | Live page verified | | Goodpods | 39 in cached page | Not exposed | Year-old search cache; live fetch blocked | | Podchaser | 62 in cached page | May 6, 2026 | Three-month-old search cache; live fetch blocked |

2/2 live-accessible directories have a catalog consistent with the feed. The other two are unverified live, so this is not a 50% directory-quality score. Apple's lookup date trails its page; the latest episode is present on the live page.

The cached Goodpods profile uses the former Galileo publisher name. Podchaser's cached profile exposes the old Anchor feed URL, but that URL successfully redirects to Transistor. The old address is therefore not a broken subscription route.

Next: check the owner view of Goodpods and Podchaser and refresh branding or profile links if they are still outdated. Keep the working feed redirect in place. This turns directory consistency into a concrete publishing checklist without penalizing a show for our access limits.

Source: Four-directory identity and catalog check; live and cached sources distinguishedFile: evidence/directories.json

Three actions for the next publishing cycle

Correct the two transcript introductions. Add a brief proper-name review to the production checklist. Verify against audio; recheck the public text after publishing.
Run the Cisco title framing as a promotion test. Compare channel-specific impressions and clicks when available. Treat any result as a hypothesis until the comparison is credible.
Clarify YouTube destinations. Confirm the two handles' channel identities and label host versus show links consistently. Check resulting subscription clicks if instrumentation exists.

For the owner review, confirm whether the priority is audience growth, sponsorship, or another business outcome. That decision should control the next set of recommendations.

What would change the recommendations?

Was the summer gap a planned season break, and what cadence is promised to listeners?
Do the host and show YouTube handles lead to the same channel or distinct destinations?
Which channel produces relevant subscribers or sponsor inquiries, and can that path be measured?
Does OP3 or the host provide tracking coverage sufficient for comparable episode windows?

Scope and confidence

Snapshot collected September 4, 2026. This is a public website and RSS audit with manual editorial judgment. The five episode pages cover August 12–September 2; peer feeds use a longer history. Transcript presence was inspected in HTML, but completeness against audio, playback, mobile interaction, search rankings, AI citations and signup completion were not tested. The sponsor-information link is present on the homepage; the sponsor destination and inquiry conversion were not audited.

The scorecard is a pilot, not a calibrated growth model. It does not grade production budget, reward missing funding tags, rank shows by catalog size, or penalize unavailable analytics. Recommendations are conditional where the evidence cannot establish a defect.

Sources

Chain of Thought public website, September 4, 2026evidence/web.json · 2026-09-04T23:10:42.465341+00:00
Reviewed homepage, start-here, subscribe, robots, sitemap index and five latest episode HTML pages. Text and structured data collected without browser interaction.
Live RSS and Apple lookup snapshot, September 4, 2026evidence/benchmark.json · 2026-09-04T23:11:31Z
Public podcast-benchmark snapshot for four AI podcasts; cadence uses 183 days, transcript percentage counts RSS tags only.
Release counts recomputed from saved RSS publication datesevidence/benchmark.json · sqlite · 2026-09-05T00:10:15.182459+00:00
Counts published RSS entries in the 183-day window. rss_episodes is loaded from the saved public RSS snapshot by build_report.py.
SQL query
```
SELECT show, COUNT(*) AS releases FROM rss_episodes WHERE julianday(published_at) >= julianday('2026-03-05T23:11:31+00:00') AND julianday(published_at) <= julianday('2026-09-04T23:11:31Z') GROUP BY show ORDER BY show
```
Transcript tag coverage in the public RSS feedsevidence/benchmark.json · sqlite · 2026-09-05T00:10:15.182459+00:00
Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.
SQL query
```
SELECT show, COUNT(*) AS items, SUM(transcript_tag) AS tagged, AVG(transcript_tag) AS coverage FROM rss_episodes GROUP BY show ORDER BY coverage DESC, show
```

Calendar days between COT releasesevidence/benchmark.json · sqlite · 2026-09-05T00:10:15.182459+00:00

Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.

SQL query

WITH recent AS (SELECT date(published_at) AS published FROM rss_episodes WHERE show='Chain of Thought' ORDER BY published_at DESC LIMIT 10), gaps AS (SELECT published, CAST(julianday(published)-julianday(LAG(published) OVER (ORDER BY published)) AS INTEGER) AS days FROM recent) SELECT published, days FROM gaps WHERE days IS NOT NULL ORDER BY published

Five reviewed episode pagesevidence/web.json · sqlite · 2026-09-05T00:10:15.182459+00:00
Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.
SQL query
```
SELECT COUNT(*) AS pages, SUM(transcript) AS transcripts, SUM(chapters) AS chapters, SUM(name_issue) AS flagged FROM reviewed_pages
```
Publisher-reported homepage countersevidence/web.json · sqlite · 2026-09-05T00:10:15.182459+00:00
Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.
SQL query
```
SELECT metric, value, qualifier FROM publisher_metrics ORDER BY metric
```
Public Listen Notes listing observationsevidence/listennotes.json · sqlite · 2026-09-05T00:10:15.182459+00:00
Computed from the saved public evidence by build_report.py. Name issues count confirmed variants only; other introductions are not certified error-free.
SQL query
```
SELECT show, score, global_band, listing_url FROM listen_notes_scores ORDER BY score DESC, show
```
OP3 lookup evidenceevidence/op3.json
Independent Podcasting 2.0 RSS tag check, September 4, 2026evidence/podcasting2.json
Six page-foundation checks on five recent episode pagesevidence/technical-seo.json · sqlite · 2026-09-05T00:10:15.195394+00:00
Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.
SQL query
```
SELECT SUM(passed) AS passed, COUNT(*) AS assessed, 100.0 * SUM(passed) / COUNT(*) AS score FROM page_checks
```
Fixed eight-query public search panel and inspected link examplesevidence/visibility.json · sqlite · 2026-09-05T00:10:15.196404+00:00
Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.
SQL query
```
SELECT SUM(owned) AS owned_hits, AVG(owned) AS owned_rate, SUM(CASE WHEN exact_title=1 THEN episode ELSE 0 END) AS episode_hits FROM query_results
```
Public listener feedback, by platform and storefrontevidence/reviews.json · sqlite · 2026-09-05T00:10:15.196940+00:00
Executed over the reviewed JSON snapshot loaded into named SQLite tables by report_additions.py.
SQL query
```
SELECT rating, ratings, sample FROM apple_feedback;
SELECT theme, COUNT(DISTINCT review_id) AS reviews FROM review_coding GROUP BY theme ORDER BY theme
```
Four-directory identity and catalog check; live and cached sources distinguishedevidence/directories.json