Go-to-Market and SEO/AEO

What to fix first on your site and in your go-to-market: Search Console and Bing setup, SEO and agent-readiness checklists, how to measure AI visibility honestly, what the top agentic-traffic companies actually did, and skills your AI can run. Every claim is dated and labeled by its evidence.

Get this Base for your AI

Use its knowledge base and reusable skills in your own AI conversations.

The knowledge base1078 cited facts from 88 documents, plus how they connect and when.

Skills

Add structured dataWrites JSON-LD structured data for one page from text already visible on that page, using only types Google still supports, and hands the patch to an engineer with validation notes. Use when a page type lacks markup, or when the person says add schema, add structured data, add JSON-LD, why no rich results, should we add FAQ schema or mark up this page. Refuses retired types such as FAQPage and HowTo for rich results, and never adds a rating or review that is not genuine. States plainly that markup serves rich results and that tests found no lift in AI citations on pages already cited. An engineer reviews before deploy. Not for a site-wide review, which is gtm-audit-site, or for changing the words on the page, which is gtm-refresh-page.

Audit the siteAudits a company's own website for search and for AI agents against two dated checklists, recording pass, fail, not-applicable or not-checked with evidence for each check, then lists what to fix first. Use on a new site, after a redesign, a move or a CDN change, or when the person says audit our site, why are we not showing up, are we blocking AI crawlers, can ChatGPT read our site, check our SEO, what should we fix first or is our site ready for agents. Starts with crawl access and rendering, treats llms.txt and Markdown as optional, and tracks failed checks and never a scanner score. Runs only read-only commands on the owner's own domain and changes nothing on the site. The owner chooses which fixes go to engineering. Not for checking page wording, which is gtm-review-claims, or for measuring whether a fix worked, which is gtm-measure-change.

Check freshnessChecks whether this copy of the template's guidance is still current. Lists files past their recheck date, re-reads the primary URLs in ATTRIBUTION.csv, reads the public master's CHANGELOG.md through its share link, and reports the differences on a thread. Use monthly, before relying on a volatile file, or when the person says is this still current, is this guidance out of date, what changed, anything new from Google, update this template or how old is this copy. A copy is a one-time fork that receives no updates, so this skill is how it catches up. Edits nothing until a person accepts each difference. Not for pages on the company's own website, which is gtm-refresh-page.

Draft an answer pageDrafts one answer-first page for one real question a buyer asks, using only verified claims and the company's own documents, and registers it at draft. Use when a prompt in the baseline has no page that answers it, or when the person says write a page about this, we need a page for this question, write a blog post, turn this into a page, draft an article or help us show up for this question. Puts a two-sentence direct answer first, gives each sub-question a labeled heading, lists anything unverified as a TODO and not as text, and needs something only this company can say. Drafts one page per run and refuses bulk generation. A named reviewer decides, after gtm-review-claims. Not for a page that names a competitor, which is gtm-draft-comparison-page, or for updating a page that already exists, which is gtm-refresh-page.

11 more skills

Draft a comparison pageDrafts one fair comparison or alternatives page against one real competitor, stating only what that competitor documents, with its source and the date read, and states where the competitor is the better fit. Use when a competitor keeps appearing in AI answers or sales calls, or when the person says us vs them, compare us to a competitor, write an alternatives page, why choose us over them, build a versus page or they rank for our name. Every competitor statement becomes a claim with a verbatim quote from the competitor's own page. Silence on their site is never treated as a gap. Drafts one competitor per run, for a person to edit, and refuses bulk generation. The owner approves every competitor statement, with a legal check if the company has a rule. Not for a general question with no competitor named, which is gtm-draft-answer-page.
Draft a release postDrafts a release post and a short changelog entry from work that has actually shipped, with evidence for each change and a claim behind each number. Use when a version ships, or when the person says write the release notes, announce this feature, what did we ship, write a launch post, draft the changelog or we shipped something. Reads done tasks and the company's changelog, separates user-visible changes from internal work, and suggests where to announce with tagged links. Never announces work that has not shipped, and never makes a performance claim without a measurement. The owner approves the wording and a person publishes. Not for an evergreen page, which is gtm-draft-answer-page, or for a monthly summary, which is gtm-write-monthly-report.
Measure a changeLogs a website or content change before it goes live, then judges afterward whether it worked, using control prompts, control pages and intervals, and reports the result or "no detectable change". Use before publishing anything meant to move search or AI visibility, and again after the wait, or when the person says did that work, did our rewrite help, prove this change worked, are we getting cited more, was it the new page or how do we test this. Requires a log entry first, a baseline, a crawl of the page and then at least two more weeks. Computes treated change minus control change, and keeps nulls. The owner accepts the reading before anything is claimed publicly. Not for the regular monthly numbers, which are gtm-run-ai-baseline and gtm-write-monthly-report.
Position the productWrites a company's positioning with its owner and starts the claims file. Lists the real alternatives customers use, the attributes only this product has, the value they bring, the best-fit customer and a market category, then writes the canvas and three one-sentence options. Use when the positioning file is empty, when the website says different things on different pages, or when the person says who are we for, what makes us different, how should we describe ourselves, sharpen our messaging, we sound like everyone else or what category are we in. Every attribute ties to a source and every factual statement becomes a draft row in the claims file. The owner picks the category and best-fit customer and approves the canvas. Not for writing a page, which is gtm-draft-answer-page, or a head-to-head page, which is gtm-draft-comparison-page.
Prepare a directory listingPrepares a submission packet for one agent or connector directory, such as Claude, ChatGPT, Cursor, the MCP Registry or skills.sh, checking the product against that directory's current requirements. For products with a real API, MCP server, plugin or skill only. Use when a connector or plugin is ready to list, or when the person says list us in the Claude directory, submit our MCP server, get into the ChatGPT apps, publish our plugin, how do we get listed or check our listing status. Reads each directory's rows, writes listing text within its limits, and tracks status and feedback afterward. An organization admin submits. This skill never submits and never enters credentials. No company reports what share of signups a directory brings, so it promises none. Not for a website review, which is gtm-audit-site.
Refresh a pageFinds pages on the company's site that are due for review, re-reads the source behind each claim, and proposes edits as a diff, with claim status changes. Use when review_due has passed or a claim has expired, or when the person says this page is out of date, refresh our old posts, update this article, traffic dropped on this page, which pages are stale or can we delete this page. Dates change only when content changes, and pages that still earn citations are not removed without a person's decision. Handles up to three pages per run. The owner approves every edit. Not for a brand new page, which is gtm-draft-answer-page, or for asking whether this template's own guidance is current, which is gtm-check-freshness.
Review claimsChecks a draft before anything is published by matching every factual sentence to the claims file, re-reading each source and comparing the quote, then reporting a pass or fail table on the thread. Use before publishing any page, post, listing or email, or when the person says fact check this, review this draft, can we say this, is this accurate, check the sources, is this safe to publish or check our claims. Flags unsupported statements, expired claims, quotes that no longer appear, vendor numbers presented as independent, superlatives and banned phrases. Comments only and never edits the draft. The reviewer decides, and a fail blocks moving the task to done. Not for writing the draft, which is gtm-draft-answer-page or gtm-draft-comparison-page, or for updating an old page, which is gtm-refresh-page.
Run the AI visibility baselineBuilds and runs a company's monthly AI visibility baseline. Sets up a prompt list with paraphrases and control prompts, plans enough runs per engine, turns a person's results into run rows and reports a rate with its run count and interval, never a rank. Use before content work and then monthly, or when the person says does ChatGPT mention us, are we showing up in AI answers, what do assistants say about us, track our AI visibility, how do we compare in Perplexity or run the prompts again. The floor is 5 runs per prompt per engine, and a monthly rate needs 60 pooled runs per prompt group on one engine. The owner approves the prompt list and chooses which gaps to act on. Not for judging whether one specific change worked, which is gtm-measure-change, or for search and analytics setup, which is gtm-set-up-measurement.
Set up measurementWalks a person through setting up Google Search Console, Bing Webmaster Tools, analytics for AI assistants and a signup question, and reports what is done and what is missing. Use before any content work, when no property is verified, when the person cannot see AI traffic, or when they say set up Search Console, connect analytics, are we tracking ChatGPT traffic, how do I see AI referrals, add a how-did-they-find-us question or why is AI traffic not showing up. A person signs in, grants access and approves any change to the signup form. This skill never receives credentials. Starts from Google Analytics' built-in AI Assistant channel and treats custom channel groups as optional. Not for measuring brand mentions in AI answers, which is gtm-run-ai-baseline, or for judging a change, which is gtm-measure-change.
Start hereSets up a fresh copy of the Go-to-Market and SEO/AEO template for one company. Records the company's site, audience and goal, confirms which checks apply, creates the first month's task threads and shows how to paste the standing rules into AGENTS.md or CLAUDE.md. Use first, right after copying the template, or when the person says get started, set this up, I just copied this, what do I do first, where do I begin or walk me through this. Threads do not travel with a copy, so this skill is what creates the work list. Stops for the owner to confirm the domain, site type and profile before any thread exists. Not for the work itself, which each task hands to its own skill.
Write the monthly reportWrites the monthly go-to-market report from the company's own measurement tables, with a run count beside every rate, referral signups kept apart from activated users, and three proposed actions with reasons. Use in the first week of each month, or when the person says monthly report, how did we do this month, what should we do next, growth update, what changed or give me the numbers. Covers channels, AI visibility, search, pages due, directory installs and open tests, and states the data gaps and the known breaks in each series. Reports traffic as traffic and never as success. The owner chooses the actions. Not for running the prompts, which is gtm-run-ai-baseline, or for deciding whether one change worked, which is gtm-measure-change.

1078 facts · 15 skills · 88 source documents

Selected findings

Tested AEO tactics including llms.txt, JSON-LD on already-cited pages, and adding a year in the title produced null or inconclusive results in live-engine testing, with no stable cross-platform causal effect found.

playbook-gtm · ¶9

Google's AI optimization guide states llms.txt is unnecessary for Google Search and neither helps nor harms visibility, though keeping it for other AI systems is fine.

evidence-what-change · ¶1–5 (2)

Google Search Console's Generative AI performance report tracks impressions by page, country, device, and date, but does not separate AI Overviews from AI Mode.

evidence-engines · evidence-engines.md [¶1–6], Google: AI Overviews and AI Mode — Gaps +1

Google's dynamic rendering is explicitly not recommended and should only be used as a last-resort workaround, with server-side, static, or client-side rendering with hydration preferred instead.

Google Search - Docu · ¶775–776

A 4.56 MB documentation index can leave an agent able to consume only about 2% of its contents, since files under roughly 100 KB are what current agents can reliably process in full.

Standards - Agent-Fr · ¶62–67

You canDownload filesUse with your AIMake a copy

How it connects

Files88 files · 1.2 MB

MDGoogle Search - Canonicalization.md4 KBDownload

MDGoogle Search - Specify a Canonical URL.md16 KBDownload

MDGoogle Search - Redirects.md12 KBDownload

MDGoogle Search - Localized Versions (hreflang).md19 KBDownload

MDGoogle Search - JavaScript SEO Basics.md17 KBDownload

MDGoogle Search - Pagination and Incremental Loading.md8 KBDownload

MDGoogle Crawling - Faceted Navigation.md6 KBDownload

MDGoogle Crawling - HTTP Status Codes.md10 KBDownload

Download all · 1.2 MB

For AI assistants

This page is a Base: a knowledge base on Base (createbases.com) in which every fact is tied to the passage it came from. It shows 5 of the 1078 facts in “Go-to-Market and SEO/AEO”, from 88 documents.

Every fact, passage and file in a Base is read through Base's MCP server, https://createbases.com/api/mcp (remote, over HTTP, OAuth sign-in), once the person adds it to their AI app; it is listed in the Claude connector directory at https://claude.ai/directory/createbases. Through it, open_shared_base opens this Base from https://createbases.com/s/u3vdj19fl2s5rys11zcay5hxiz, explore_base answers a question with the passages behind the answer, and download_file delivers its files.

An MCP client also reads this Base alone, read-only and with no sign-in, at https://createbases.com/api/mcp/u3vdj19fl2s5rys11zcay5hxiz.

The download links on this page are private to this share and open in a browser.