Corpus Plan

a free website audit, then a workspace for what to fix first. pro adds drafts you review and post yourself.

[Corpus Plan @Corpus_hq]

@wespreadjam @wespreadjam The cite-vs-repeat gap is the part I'd want more on: cited in 35% of answers, invented details in only 10%. Did that gap shrink over time, or stay flat until the articles were deleted?

[Corpus Plan @Corpus_hq]

Four numbers treated as one: mentions, citations, recommendations, and clicks. Business Hub finally shows clicks. Are you counting all four?

[Corpus Plan @Corpus_hq]

@neilpatel If that happens, the content calendar turns into an 'answer calendar': which questions do buyers ask, and is your brand the answer? What's the 4th-place channel in the forecast?

[Corpus Plan @Corpus_hq]

@TheoAGI "Read the findings. Ignore the number." That line should be on every audit. I'd add one check: if your offer and price only appear after JavaScript runs, the crawler can get a 200 and still have a blank page.

[Corpus Plan @Corpus_hq]

@lilyraynyc Same-day core update plus first-party ChatGPT click data is a messy experiment. I'd split GSC branded vs non-branded before blaming the update, and treat Business Hub as one engine's view, not the whole market.

[Corpus Plan @Corpus_hq]

@Kevin_Indig The 28-day mention and click window is a real step. The gap still matters: if brand-name-only answers aren't counted, and prompts aren't shown, you're seeing the wins, not the questions you lost.

[Corpus Plan @Corpus_hq]

Steal this 30-second check before you pay anyone for "AI visibility":

1. Open https://t.co/5IhkK9I9DG
2. Search for OAI-SearchBot, PerplexityBot, Claude-SearchBot
3. If "Disallow: /" sits under them, ChatGPT, Perplexity and Claude search can't read your site
4. GPTBot and ClaudeBot are training bots. Blocking those is a separate choice

A lot of "block AI" templates block both. How many of you have been bitten by this?

[Corpus Plan @Corpus_hq]

@OpenAI The gap I'm watching is how a mention gets counted. If it only records cases where the website is cited, brand-name-only answers won't show up, and that's often the whole job in professional services.

[Corpus Plan @Corpus_hq]

The method, from Glenn's post: Search Console, last 6 months, filter to France, then look at the Performance report around July 22 and the Generative AI impressions. Split branded vs non-branded before you blame AI.

[Corpus Plan @Corpus_hq]

Glenn Gabe's France AI Overviews cut is the cleanest before/after we've got (rollout July 22, 2026). Some sites lost a lot of clicks while impressions held; some barely moved.

If you have France traffic: have you pulled that Search Console split yet, and which way did you fall?

[Corpus Plan @Corpus_hq]

The $4 UGC video is not the hard part. Choosing which 10 of the 100 ads deserve a budget, before the pixel has a week of data, is. How are you picking winners right now?

[Corpus Plan @Corpus_hq]

The weekend's most agreed-with hot take was "Marketing (even with AI) is still very hard."

I'll go further: AI made the easy 80% faster and the hard 20% more obvious.

What's still hard for you — the offer, the channel, or knowing if it worked?

[Corpus Plan @Corpus_hq]

Be honest: have you ever blocked AI bots in robots.txt with a "protect my content" template, then realized you also blocked the search crawlers?

No judgment. I just want a show of hands.

[Corpus Plan @Corpus_hq]

@CabotTCroll Agreed. AI shrunk the time from "blank page" to "first draft." It did not shrink the time from "first draft" to "someone cares." Curious which part is still hardest here: the offer, the hook, or the distribution.

[Corpus Plan @Corpus_hq]

@gregisenberg The per-deliverable shift only works if clients can tell a good deliverable from a fast one. Otherwise the race is to cheaper packets, not better work. Are buyers already asking for the raw prompt, or just the file?

[Corpus Plan @Corpus_hq]

Corpus Plan's audit checks this. It fetches your homepage's other version (www or bare) and flags it when both load the same page, neither redirects, and no HTML canonical picks one.

Free preview, no account or card: https://t.co/p266Uo8vFG
Full report: free account.

[Corpus Plan @Corpus_hq]

One homepage or two?

1. Load your site with www. and without.
2. Do both end on one address?
3. Both stay put? View Source: does rel="canonical" name one?
4. If not, redirect one to the other.

Ours, Nov 4: www redirected to the bare domain. One address, not a ranking fix.

[Corpus Plan @Corpus_hq]

Nov 3, 3 recorded checks: the Organization logo URL in Calendly's homepage markup returned 404 as served to us.

Check your homepage JSON-LD for "logo", then open that URL. Does an image load? We did not check search results.

[Corpus Plan @Corpus_hq]

On Nov 2, one run: 42 of 50 hand-picked software sites served us their homepage. 1 og:image came as a 9,397,453-byte animated GIF when we didn't ask for WebP, and a 1,847,224-byte WebP when we did.

X's developer docs (archived Feb 2026): "Images must be less than 5MB in size."

[Corpus Plan @Corpus_hq]

Want help checking page signals? Corpus Plan includes title and heading checks. Try the free preview without an account: https://t.co/p266Uo8vFG

The full report and audit-based action plan require a free account.

[Corpus Plan @Corpus_hq]

Homepage title check: open your homepage, View Page Source, then find <title>. Read it alone: does it describe the page, or just say "Home"?

Our example: "Website Audit Tool: SEO, Speed, AI Visibility" (title excerpt, read Nov 3). This checks clarity, not rankings.