PART FOUR — DOING IT
Chapter 14 — The First Ninety Days
Everything in the book, sequenced into ninety days with dependencies running downward: entity before corroboration, reachability before content, measurement before any claim of success. Days 1–30 find out what is true, days 31–60 fix it and build the baseline, days 61–90 build the things that compound.
From Becoming the Answer by Jeremy Osborn · 841 words
Everything in this book, sequenced. Dependencies run downward: entity before corroboration, reachability before content, measurement before any claim of success.
Days 1–30: find out what’s true
The goal of the first month is to know where you stand and to fix what’s silently broken.
| # | Action | Owner | Effort |
|---|---|---|---|
| 1 | Write the canonical facts page; get positioning sign-off | Brand | 1 day |
| 2 | Audit every external property against it; count contradictions | Brand / SEO | 1 week |
| 3 | Check robots.txt against the bot table — verify, don’t add groups | Engineering | 1 day |
| 4 | Audit Cloudflare and WAF settings — deadline 15 September 2026 | Engineering | 1 day |
| 5 | View-source and curl tests: is key content in the initial HTML? | Engineering | 2 days |
| 6 | Pull AI-bot server logs; quantify 404 and redirect waste | Engineering | 2 days |
| 7 | Turn on Search Console generative AI reports and Bing AI Performance | SEO | 1 hour |
| 8 | Verify the GA4 AI Assistant channel; document the Direct gap | Analytics | 1 day |
| 9 | Harvest real customer language from sales calls, support, site search | Content / Sales | 1 week |
| 10 | Baseline accuracy audit: read fifty AI answers about your brand | Marketing | 2 days |
| 11 | Brief legal on the review rule, AI Comply, and EU transparency duties | Legal | 1 day |
| 12 | Name one accountable owner; schedule the quarterly forum | CMO | 1 day |
What you should have at the end: a one-page diagnostic — number of entity contradictions, percentage of key content server-rendered, AI crawler 404 rate, accuracy findings, and whether any platform is currently blocked by misconfiguration.
That last item justifies the whole month more often than not. Blocking OAI-SearchBot while intending to block GPTBot is common, invisible, and entirely self-inflicted.
Days 31–60: fix it and build the baseline
| # | Action | Owner | Effort |
|---|---|---|---|
| 13 | Fix entity contradictions: LinkedIn, Crunchbase, directories, boilerplate | Brand / Comms | 3 weeks |
| 14 | Add a plain-text identity statement to the homepage or About page | Brand | 1 day |
| 15 | Ship Organization schema with a complete sameAs | Engineering | 1 day |
| 16 | Create a referenced Wikidata item; record the Q-ID | SEO | 1 day |
| 17 | Move client-side-only content to server-side rendering | Engineering | 2–4 weeks |
| 18 | Fix the worst AI-crawler 404s and redirect chains | Engineering | 3 days |
| 19 | Build the 120–200 question set from the month-one harvest | Content | 1 week |
| 20 | Buy one synthetic-prompt tool and one real-prompt tool; ask both for run counts | SEO | 2 weeks |
| 21 | Map the third-party roundups cited for your top ten commercial queries | PR / SEO | 1 week |
| 22 | Correct factual errors about you in those sources | PR | 2 weeks |
| 23 | Audit the product feed; populate Product Highlight and Product Detail | Ecommerce | 2 weeks |
| 24 | Review bot mitigation against agentic browser traffic | Security / Eng | 1 day |
What you should have: a measurement baseline with rolling mention rate and citation rate reported separately, and a corrected entity footprint.
Do not report a trend yet. You don’t have one. Reporting a trend from three weeks of data is exactly the error Chapter 11 warns about, and doing it once sets a precedent you’ll spend a year unwinding.
Days 61–90: build the things that compound
| # | Action | Owner | Effort |
|---|---|---|---|
| 25 | Publish constrained-recommendation pages for the top twenty intersections | Content | 6 weeks |
| 26 | Publish prices and specifications wherever commercially possible | Content / Product | 2 weeks |
| 27 | Commission the first original research study | Marketing | starts now |
| 28 | Launch digital PR aimed at the outlets that actually get cited | PR | ongoing |
| 29 | Establish a named, disclosed presence on the two key forums | Community | ongoing |
| 30 | Launch a sentiment-neutral review solicitation program | CX | ongoing |
| 31 | Fund a YouTube pilot with real substance | Content | ongoing |
| 32 | Migrate FAQ content into feed Q&A attributes; attach document links | Ecommerce | 2 weeks |
| 33 | Publish /.well-known/ucp if you sell where Universal Cart is live | Engineering | 1 sprint |
| 34 | Run the first quarterly agent-completability test | Ecommerce | 1 day |
| 35 | Put content on a refresh cycle | Content | ongoing |
| 36 | First cross-functional forum: review inputs and outputs together | CMO | half day |
What you should have: a report with four output metrics on rolling averages, five input metrics you control, an accuracy finding, and an explicit statement of what you still cannot measure.
That last section is the most credible thing in the document. It’s also what protects you when someone asks why the numbers moved and the honest answer is that they didn’t move — they varied.
What not to do in the first ninety days
Don’t restructure the content library for chunkability
Don’t commission a schema program aimed at AI citations
Don’t pay anyone for llms.txt
Don’t block Google-Extended expecting it to affect AI Overviews
Don’t press the Search Console AI opt-out
Don’t seed reviews, forum posts or testimonials — ever
Don’t report a trend before you have properly sampled data
Don’t fund an “agentic commerce transformation.” Fund a feed and a manifest.
If you do nothing else
Check your crawler configuration. Make your identity consistent everywhere it appears. Put your real content in the initial HTML. Start measuring properly.
Four things, all graded A, all cheap, all undone in most organizations right now.
Everything else in this book is optimization on top of them.