type.com team guide
AI for SEO: a practical guide
Set up one SEO Space that reviews Search Console, briefs from real data, triages technical issues, and audits AI search, while people publish.
How should an SEO team use AI?
An SEO team should use AI to assemble and analyze, meaning Search Console reviews, briefs built from real query data, decay and technical triage, internal links, and AI search audits, while people own the facts, the expertise, and every publish decision.
This guide is for SEO leads and content marketers at B2B and ecommerce companies. By the end, seven workflows run in one shared type.com Space, each with paste-ready instructions, an example, a review rule, and failure modes.
SEO sits inside a bigger marketing loop of positioning, paid media, and newsletters. That loop is covered in AI for marketing teams. This guide goes deeper on search: what to automate, what stays human, and how to stay within Google's guidance.
The running example is Copperline, a fictional company selling accounts payable software to finance teams. Nadia leads SEO, Theo leads content, Ravi ships technical fixes, and Jess runs marketing. Ecommerce notes appear where workflows differ.
| Workflow | Channel | Runs (Eastern) | Time replaced |
|---|---|---|---|
| 1. Weekly Search Console review | search-performance | Mon 9:00 AM | 1–3 h a week |
| 2. Content briefs from query and SERP data | content | On request | 1–2 h per brief |
| 3. Content refresh and decay detection | content | First Tue 10:00 AM | 2–4 h a month |
| 4. Technical audit triage | technical-seo | Thu 10:00 AM | 1–3 h a week |
| 5. Internal linking suggestions | content | When a post publishes | 20–40 min per page |
| 6. AI search visibility audit | ai-search | Second Wed 10:00 AM | 3–5 h a month |
| 7. Monthly SEO report | search-performance | First Mon 8:00 AM | 3–6 h a month |
What does Google say about using AI for SEO content?
Google allows AI assistance, but treats generating many pages without adding value as scaled content abuse and asks you to fact-check all AI-generated content, metadata included, before publishing.
Google's guidance on using generative AI content says AI is useful for research and for structuring original content, but warns that models predict likely words rather than retrieve facts. It asks site owners to fact-check all AI-generated content before publishing, including titles, meta descriptions, structured data, and alt text.
The scaled content abuse policy describes many pages generated mainly to manipulate rankings, "no matter how it's created," and names using generative AI to produce many pages without adding value as an example. Google's helpful content guidance lists "extensive automation to produce content on many topics" as a warning sign and asks whether content shows first-hand expertise.
So the rule here is simple: AI drafts and analyzes; a named person owns every fact, adds the experience only your company has, and presses publish. That first-hand detail is also what earns links, rankings, and citations.
- AI may: pull and compare data, read ranking pages, draft briefs, outlines, titles, refresh lists, tickets, and reports.
- People must: supply expertise and examples, verify every number and claim, approve metadata and schema, and publish.
- Never: publish pages in bulk from a keyword list, invent statistics or quotes, or let an automation edit the live site.
How do you set up an SEO Space in 30 minutes?
Create a Space that runs Claude, add read-only search, analytics, and CRM connections, create four channels, and paste starter instructions; it takes about 30 minutes.
If you already have a Marketing Space, add these channels there; a separate SEO Space suits a dedicated team or agency with different access.
In type.com, AI acts with the effective access of the person who started the work, and a connection works only in the Spaces it is assigned to. A Read-only setting tells the agent not to make changes but does not block them at the service, so set the real limit in Google and your CMS.
- 1
Create the Space
Choose Create, then Space, and pick the Marketing template, which covers SEO. Edit its prompt. The template only proposes skills, automations, and connections; you approve each.
- 2
Choose Claude
Claude is the documented default for research and writing. Pick Codex if your SEO work is mostly site code.
- 3
Add connections with narrow access
Add the tools below in Space settings, Connections. For Search Console, use a catalog connection, a Custom API with your own Google OAuth app and the webmasters.readonly scope, or CSV exports in Drive.
- 4
Create four channels
search-performance (reviews, reports), content (briefs, refreshes, links), technical-seo (crawl and indexing), and ai-search (AI visibility).
- 5
Map Slack if the team lives there
Invite the type.com app to #seo and map it to search-performance with Mentions only.
- 6
Write instructions and definitions
Paste the starter below into Space settings, Details, Instructions; it is sent with every message. Then have AI draft an SEO definitions document: brand terms, money pages, conversions, a do-not-link list, and approvers. Memory is read-only, so exact rules belong here.
Starter Space instructions
You support Copperline's SEO and content team. Copperline sells accounts payable automation software to finance teams at companies with 200 to 2,000 employees. Owners - Nadia: SEO lead. Owns the weekly review, technical triage, the AI search audit, and the monthly report. - Theo: content lead. Owns briefs, refreshes, internal links, and final edits. - Ravi: web developer. Ships technical fixes. - Subject experts supply first-hand detail and review claims. Definitions (full version: the "SEO definitions" document) - An organic conversion is a demo request or trial signup whose original source in HubSpot is organic search. - Brand queries contain "copperline"; report them separately. - Skip the 3 most recent days of Search Console data; they are incomplete. - Times are Eastern. How to work 1. Analyze and draft. Never publish, edit the site, submit URLs or sitemaps, or change robots.txt or Search Console settings. 2. Give the source and dates for every number. If data is missing, say so and stop. Never estimate it. 3. Label inferences "inference" and projections "est." 4. Never invent statistics, quotes, features, or first-hand experience. Write [NEEDS EXPERT] with a question instead. 5. Never propose many near-identical pages, such as one per city or keyword variant. 6. Put briefs, audits, and reports in documents, and name who approves each change.
| Connection | Start with | Why | Later |
|---|---|---|---|
| Google Search Console | webmasters.readonly scope, or CSV exports | Queries, pages, clicks, CTR, position | Keep read-only; people submit URLs |
| Google Analytics (GA4) | A user with Viewer access | Organic landing-page engagement | None; analysis stays read-only |
| HubSpot (or your CRM) | Organization connection, read | Demo requests and pipeline from organic | None for SEO work |
| Google Drive | Crawl exports and briefs folders only | Crawler CSVs, Page indexing exports | A drafts folder AI may write to |
| Keyword or SERP tool | Custom API with a read-only key, or exports | Top results, volumes, competitors | None |
| Slack | #seo, Mentions only | Quick questions where the team works | Channel messages, once trusted |
How do you run a weekly Search Console review with AI?
Run a Monday automation that compares complete Search Console periods, explains every lost click by position, impressions, or click-through rate, and ends with three owned actions.
When it runs: every Monday at 9:00 AM Eastern in search-performance. Nadia owns it.
Inputs: Search Console (Web search type). Google counts AI Overviews and AI Mode traffic inside the Web totals; there is no separate AI line.
Example output: on Monday, Oct 5, /guides/three-way-match had lost 640 clicks over 28 days while position held at 3.0 and click-through rate fell from 8.4% to 5.1%. The review did not guess why. Nadia checked by hand and found a new AI Overview above the results, so Theo rewrote the opening to answer in two sentences and added a worked example.
Review rule: AI never changes Search Console or the site. Nadia picks the actions, each with an owner by noon.
Shortcut: the Search Console SEO Analyzer skill in the type.com Skills Library packages this comparison as a tested, read-only script: declining pages, CTR gaps, rising queries, cannibalization, and page-1 opportunities, with projections labelled as estimates.
Weekly Search Console review: automation instructions
Every Monday, run the weekly Search Console review for Copperline. 1. Use the Web search type. Skip the 3 most recent days. Compare the latest 7 complete days with the 7 before, and the latest 28 with the 28 before. State the exact dates. 2. Report clicks, impressions, CTR, and average position, brand and non-brand separately. 3. List the 5 pages that lost the most clicks over 28 days. For each, say whether position, impressions, or CTR changed most, with before and after numbers. If position held but CTR fell by a third or more, write "check the results page by hand" instead of guessing the cause. 4. List up to 5 queries that rose 3+ positions and now rank 5 to 15 with 300+ impressions. 5. Flag queries in positions 1 to 5 with less than half our average CTR at that position. 6. Flag queries where two or more of our pages each get 20%+ of impressions. 7. End with at most three actions, each with an owner and its evidence. 8. If any day is missing or a total is zero, say so and stop. Do not change anything in Search Console or on the site.
- Reading the last two days as a drop: Search Console fills in late, so the instructions skip the newest three days.
- Explaining a CTR drop without seeing the page: the review flags it for a human SERP check instead of inventing a reason.
How do you create content briefs from real SERP and query data?
Have AI build the brief from your Search Console queries and the pages that rank today, with marked gaps for a subject expert, so the page starts from evidence and adds something competitors lack.
When it runs: on request in content, whenever Theo or a writer starts a page. Theo owns it.
Inputs: Search Console queries, the top 10 organic results for the target query (pasted, or from a SERP or keyword tool export), and the Copperline sitemap.
Example output: the brief for "accounts payable KPIs" found 1,140 impressions across three Copperline pages, none ranking above 24, so a new page made sense. Seven of the top 10 results were KPI guides, so the outline led with definitions and formulas. Two [NEEDS EXPERT] questions went to Dana in finance operations, who supplied cost-per-invoice figures from Copperline's customer benchmarks.
Review rule: Theo approves the outline, and nothing is drafted until the expert questions are answered.
Content brief: skill instructions
Write a content brief for the target query I give you. 1. From Search Console, list queries (last 90 days, 10+ impressions) that already reach our pages on this topic, and which page gets them. If one of our pages ranks in the top 20, recommend updating it instead and stop for approval. 2. Read the top 10 organic results I paste. For each, note its format (guide, tool, template, comparison, product page), its headings, and anything useful the others lack. If a page will not load, say so. 3. Say what searchers most likely need, based on the formats that rank. Mark it "inference". 4. Draft a title under 60 characters, a meta description under 155, an H1, and an outline of H2s and H3s, with one line on what each section must answer. 5. List what only Copperline can add (first-hand data, customer examples, screenshots, expert opinion) as [NEEDS EXPERT] questions for a named expert. Do not answer them yourself. 6. Suggest 3 to 5 existing Copperline pages to link from and to, with anchor text. 7. List every fact the writer must verify, with the source you found. 8. Save it as a document titled "Brief – <query>". Do not write the article.
- Briefing a page you already have: step 1 checks for an existing ranking page first, which also prevents cannibalization.
- Copying the top results' outline: the expert questions force something new into the page.
How can AI detect content decay and plan refreshes?
Once a month, have AI compare each page's clicks with last year and the previous period, flag pages that fell on both, and recommend refresh, merge, redirect, or leave alone for a person to decide.
When it runs: the first Tuesday of each month at 10:00 AM Eastern in content. Theo owns it.
Inputs: 15 months of Search Console page data and the live pages.
Example output: in September, /guides/vendor-onboarding-checklist was down 38% on last year and 31% on the prior three months, position had slipped from 4.0 to 7.2, and it quoted a 2022 survey. Theo approved a refresh, and an expert added Copperline's current onboarding steps instead of a newer statistic.
Review rule: Theo picks up to three pages a month. Merges and redirects also need Nadia.
For ecommerce: product page decay is often a stock or feed problem, not a copy one. If you sell through Google Shopping, the Product Feed Auditor skill checks a Merchant Center feed export for missing attributes, policy problems, broken links, and price mismatches, by item id.
Monthly decay scan: automation instructions
On the first Tuesday of each month, find decaying content. 1. Compare each page's clicks in the last 3 complete months with the same months a year earlier and with the 3 months before. Include pages with 300+ clicks in the earlier period. 2. Flag a page as decaying if clicks fell 25% or more on both comparisons. List pages that fell only against the previous 3 months separately as "possibly seasonal". 3. For each decaying page, show the change in position, impressions, and CTR, and its 5 queries that lost the most clicks. 4. Quote anything dated on the page: years, prices, product names, screenshots, broken links, or statistics over 2 years old. 5. Recommend one of: refresh, merge into another page (name it), redirect, or leave alone, with a one-sentence reason. 6. Rank by clicks lost, stop at 10, and ask Theo to choose up to 3. Do not edit any page.
- Refreshing by changing the date: Google's guidance warns against cosmetic updates; the scan asks what is actually out of date.
- Rewriting seasonal pages in their off-season: the two-comparison rule catches them.
How do you triage a technical SEO audit with AI?
Give AI your Page indexing and crawler exports each week, and have it group issues by type and URL pattern, rank them by business impact, and draft tickets a developer verifies.
When it runs: every Thursday at 10:00 AM Eastern in technical-seo. Nadia triages; Ravi owns the fixes.
Inputs: the Search Console Page indexing export (it is not in the API, so a person exports it weekly), a crawler export from a tool such as Screaming Frog or Sitebulb, the sitemap, and the money pages list.
Example output: on Thursday, Oct 1, the top item was twelve filtered customer-story URLs returning HTTP 200 with "No results", which Google reports as soft 404s. The ticket proposed a 404 or noindex for empty filter states, verified with URL Inspection after the fix.
Review rule: AI never edits robots.txt, redirects, canonicals, or sitemaps. Ravi verifies the cause first, because a wrong canonical or robots fix can remove good pages from search.
Thursday technical triage: automation instructions
Every Thursday, triage technical SEO issues for Copperline. 1. Use the newest Page indexing export and crawler export in the "Crawl exports" folder. State their dates. If either is more than 8 days old, say so and stop. 2. Group issues by type: server errors (5xx), Not found (404), Soft 404, redirect chains and loops, "Duplicate, Google chose different canonical than user", "Crawled – currently not indexed", "Discovered – currently not indexed", noindex pages listed in the sitemap, and pages blocked by robots.txt. 3. For each group, give the count, the URL pattern (for example /integrations/*?ref=), and three example URLs. 4. Severity: High if the group includes a money page or a page with organic clicks last month. Medium if it affects more than 50 URLs. Otherwise Low. 5. Give the likely cause, labelled "inference", and the check that confirms it, such as URL Inspection on one example. 6. For each High group, draft a ticket for Ravi: problem, evidence, proposed fix, and how to verify the fix. 7. Do not change robots.txt, redirects, canonical tags, or sitemaps, and do not request indexing.
- Treating every excluded URL as a problem: many exclusions are correct, so severity depends on what the pages are worth.
- Fixing the symptom: tickets include how to verify, so Ravi confirms the cause first.
How can AI suggest internal links?
Point an RSS automation at your own blog feed so every new post triggers a skill that suggests links into and out of it, with the exact sentence and anchor text, for an editor to add in the CMS.
When it runs: whenever a post appears in Copperline's blog RSS feed, which an RSS automation sends to content. Refreshed pages run by hand. Theo owns it.
Example output: for the new accounts payable KPIs guide, the skill proposed six links in, including a three-way match guide sentence that already mentioned "days payable outstanding", and four links out. Theo accepted five.
Review rule: an editor adds links by hand. Google lists internal links among the basics that also matter for its AI features.
No public library skill covers this yet, so save the instructions below as a skill. Teammates can propose improvements as suggestions the owner reviews.
SKILL.md: internal-link-suggestions
--- name: internal-link-suggestions description: Suggest internal links into and out of a new or refreshed Copperline page. Use when a post is published, a page is refreshed, or someone asks for internal links. --- # Internal link suggestions 1. Read the target page and note the 3 to 5 queries it targets (Search Console, or its brief). 2. Read the 15 most related pages in our sitemap. 3. Suggest up to 8 links into the target page: source URL, the existing sentence where the link fits, and a 2-to-6-word anchor that describes the target. Vary anchors. Never use "click here". 4. Suggest up to 5 links out of the target page to pages that help the reader next. 5. Prefer source pages with organic clicks. Skip pages that redirect, are noindexed, or appear on the "Do not link" list. 6. Never suggest two links to the same target from one page. 7. Output a table: source, sentence, anchor, target, reason. Do not edit any page.
- Exact-match anchors everywhere: the skill varies anchors and quotes the real sentence.
- Linking to redirected or noindexed URLs: step 5 skips them.
How do you audit visibility in AI search (AEO and GEO)?
Once a month, sample how ChatGPT, Perplexity, and Google's AI Mode answer your 20 most important buyer questions, have AI record who gets cited and what is wrong about you, and pair that with a crawl-access and page-readiness check.
When it runs: the second Wednesday of each month at 10:00 AM Eastern in ai-search. Nadia owns it.
Inputs: an "AI search questions" document with 20 real buyer questions (from sales calls, question queries, and support tickets), an "AI answers log" where Nadia pastes each engine's answers in the first week of the month, and robots.txt.
A person samples answers because they vary by account, place, and day, and no engine offers a full citation report. Track the trend and the wrong facts, not one score.
Google's page on AI features and your website says AI Overviews and AI Mode have no extra technical requirements: a page must be indexed and eligible for a snippet. Google-Extended governs Gemini training and grounding, not inclusion in Search.
Other engines use their own crawlers: OAI-SearchBot for ChatGPT search (GPTBot is for training), PerplexityBot for Perplexity, and Claude-SearchBot for Claude's search. You can block training without blocking search, so set each token deliberately.
Example output: September's biggest problem was not content: a CDN bot rule returned 403 to PerplexityBot. Two answers also quoted a $49-a-month price retired in 2025, traced to an old review, and one confused Copperline with Copperline Analytics. Ravi removed the rule, Theo added a plain pricing answer to /pricing, and Nadia asked the review site for an update.
Review rule: AI never edits robots.txt or CDN rules. It drafts wrong-entity fixes, such as clearer About page facts or Organization structured data, for Nadia to approve.
Shortcut: the GEO / AI Search Audit skill automates the readiness half (crawler access, citability, schema, llms.txt, rendering) but does not query AI engines or measure citations. Shopify brands can add the Shopify AEO Checker, which shows how often a brand appears against competitors when AI shopping agents search Shopify's cross-store catalog.
Monthly AI search visibility audit: automation instructions
On the second Wednesday of each month, run the AI search visibility audit. 1. Read the "AI answers log": this month's ChatGPT search, Perplexity, and Google AI Mode answers to the 20 questions in "AI search questions". If answers are missing, list them and continue. 2. For each answer, record: was Copperline mentioned, was a Copperline page cited (which URL), and which competitors were cited. 3. Quote anything wrong or outdated about Copperline (pricing, features, integrations, company facts) and the source the answer cites. 4. Flag wrong-entity confusion: answers that mix Copperline up with a similarly named company or credit us with a competitor's feature. 5. For each question where we were not cited, check our best matching page: is it indexed, is it allowed in robots.txt for Googlebot, OAI-SearchBot, Claude-SearchBot, and PerplexityBot, does it answer the question in its first two sentences, and does its structured data match the visible text? 6. Compare with last month: questions gained and lost, per engine. 7. Rank fixes by how many questions they affect. Never claim a fix will produce citations. Do not edit robots.txt, CDN settings, or any page.
- Blocking AI crawlers by accident: CDN and firewall rules can block a bot that robots.txt allows.
- Optimizing for citations while facts are wrong: fix wrong-entity and pricing errors first.
How do you automate a monthly SEO report?
Have a first-Monday automation assemble search numbers, organic pipeline, shipped work, and AI search results into a document with a five-sentence summary, and let the SEO lead add judgment before sharing.
When it runs: the first Monday of each month at 8:00 AM Eastern in search-performance. Nadia owns it; Jess reads it.
Inputs: Search Console, GA4, HubSpot, the month's content and technical-seo threads, and the latest AI search audit.
Example output: September's summary: non-brand clicks rose 6% on August, organic demo requests rose from 58 to 64, two of three refreshed pages recovered position, and the PerplexityBot block was removed. Nadia added why /pricing clicks fell (lower brand demand, not a ranking loss) before sharing it with Jess.
Review rule: Nadia checks three numbers against the source before sharing.
Shortcut: if leadership asks whether organic visitors convert, the Landing Page Conversion Analyzer compares conversion by landing page and source against each page's own history. For a weekly report pattern, see how to automate a weekly team report.
Monthly SEO report: automation instructions
On the first Monday of each month, write the SEO report for the previous calendar month. 1. Search Console (Web): clicks, impressions, CTR, and position, brand and non-brand, against the prior month and last year. 2. HubSpot: demo requests and trial signups whose original source is organic search, against the prior month. If GA4 differs from HubSpot by more than 10%, report HubSpot and note the gap. 3. The 5 pages with the largest non-brand click gains and losses, with the cause found in the weekly reviews. 4. Work shipped: briefs published, pages refreshed, tickets closed. Link each. 5. AI search: citation counts per engine from the latest audit, and any wrong facts still open. 6. Three priorities for next month, with owners. 7. Start with a five-sentence summary for Jess, every claim backed by a number. 8. Save it as a document titled "SEO report – <month>". Do not send it.
What does a weekly SEO rhythm look like?
Each routine is a scheduled or RSS automation with a fixed time, timezone, channel, and owner, and each one ends in a specific human action.
Fixed times tell the team when to look. Keep the method in a skill and the schedule in the automation, so manual and scheduled runs follow the same rules.
- 1
Choose Create, then Automation
Pick the SEO Space. The automation opens in the Library, paused.
- 2
Add instructions, channel, and schedule
Paste the prompt, choose the channel, and set day, time, and timezone.
- 3
Save, Test, then Enable
The test runs privately as you; check three numbers against the source. Only Enable starts the schedule. For internal links, add an RSS source for your blog feed instead.
| When | Workflow | Channel | Owner | Human action |
|---|---|---|---|---|
| Mon 9:00 AM | Weekly Search Console review | search-performance | Nadia | Picks up to three actions and owners by noon |
| Thu 10:00 AM | Technical triage | technical-seo | Nadia, Ravi | Ravi verifies causes and ships High tickets |
| New blog post (RSS) | Internal links | content | Theo | Editor adds approved links in the CMS |
| First Mon 8:00 AM | Monthly SEO report | search-performance | Nadia | Checks three numbers, adds context, shares |
| First Tue 10:00 AM | Decay scan | content | Theo | Chooses up to three refreshes |
| Second Wed 10:00 AM | AI search visibility audit | ai-search | Nadia | Samples answers in week one; assigns fixes |
What guardrails keep AI safe in SEO work?
Write down what AI may read, draft, do with approval, and never do in each tool, and enforce the strictest limits in Google, your CMS, and your CDN.
Know whose access each run uses. Scheduled runs use the automation creator's identity, credentials, and billing, while Test runs as whoever presses it. Create each routine from its owner's account, and recreate it if that person leaves.
Data handling: assign search, analytics, and CRM connections only to this Space, keep tokens in connections, and ask for aggregates, not contact lists. Agencies should keep each client in its own Space. Use a private thread or Sidekick for unannounced migrations.
More patterns: AI approval workflows, and for agencies, AI for agencies.
| Tool | Read | Draft | With approval | Never |
|---|---|---|---|---|
| Search Console | Queries, pages, exports | Fix lists, titles, descriptions | A person ships in the CMS | Submit URLs or sitemaps, change settings |
| GA4, HubSpot | Organic sessions, conversions | Reports and summaries | Not needed | Edit tracking, records, or contacts |
| CMS and site | Public pages, sitemap | Briefs, outlines, drafts, links | Editor publishes | Publish, edit, or bulk-create pages |
| robots.txt, CDN, redirects | Rules and responses | Tickets with evidence | Ravi changes and verifies | Change anything directly |
What does a 30-day rollout plan look like?
Turn on one or two workflows a week, starting with read-only analysis, and add drafting work only after the reports stop needing factual corrections.
Keep a correction log in search-performance. When a mistake repeats, fix the instructions or the skill, not just the output. Memory is context, not a source of truth; see shared AI memory for teams.
- 1
Week 1: setup and the weekly review
Create the Space, connections, instructions, and definitions. Test and enable the Monday review. Record baselines.
- 2
Week 2: technical triage and the decay scan
Start weekly indexing and crawl exports, enable triage, and run the decay scan by hand to tune thresholds.
- 3
Week 3: briefs and internal links
Save the brief and link instructions as skills, add the RSS source, and run two real briefs with an expert.
- 4
Week 4: AI search and the monthly report
Write the 20 questions, sample answers, run the first audit, and build the monthly report. Review the correction log before widening anything.
How do you measure whether AI is working for SEO?
Measure the hours spent on recurring SEO assembly, the accuracy of what AI produces, and the search outcomes the workflows touch, against a baseline recorded before week one.
Log a normal week's hours and last month's numbers before enabling anything. Judge accuracy and time after 30 days, search outcomes after 60 to 90.
For more skills, see the best AI skills for B2B teams and for ecommerce. Other guides in this series: AI for ecommerce and AI for sales.
| Metric | Baseline | Good after 30 to 90 days |
|---|---|---|
| Hours on recurring SEO assembly | One logged week before setup | Clearly lower, time moved to named work |
| Report accuracy | Three numbers checked weekly | No factual corrections two weeks running |
| Clicks to refreshed pages | 28 days before each refresh | Recovering on most refreshed pages after 60 days |
| High-severity indexing issues open | Count in the first triage | Falling, each fix verified with URL Inspection |
| AI answers citing you, wrong facts open | The first monthly audit | Citations stable or rising; wrong facts closing |
Frequently asked questions
Does Google penalize AI-generated content?
Not for being AI-generated. Google says generating many pages without adding value may violate its scaled content abuse policy, and that AI content, including titles, descriptions, and structured data, should be fact-checked before publishing. The risk is unreviewed volume, not the tool.
How do I get my site cited in ChatGPT, Perplexity, and Google AI Overviews?
Let each engine's search crawler reach the page (Googlebot, OAI-SearchBot, PerplexityBot, Claude-SearchBot), answer the question plainly near the top, and keep facts about your company consistent. Google says AI Overviews need nothing beyond being indexed and snippet-eligible. No one can guarantee a citation.
Is it safe to connect Google Search Console to AI?
Yes, with a narrow grant: the webmasters.readonly OAuth scope or a restricted user, assigned only to the SEO Space. In type.com, Read-only guides the agent but does not block changes at the service, so the hard limit belongs in Google.
What is the difference between SEO, AEO, and GEO?
SEO earns rankings and clicks. AEO (answer engine optimization) and GEO (generative engine optimization) are names for earning citations inside AI answers from ChatGPT, Perplexity, and Google's AI features. The foundations overlap: crawlable, indexed pages that answer questions clearly and accurately.
Can AI write SEO content at scale?
It can draft, but publishing many lightly edited pages is what Google's scaled content abuse policy describes. Use AI for briefs, outlines, and first drafts; an expert adds first-hand detail and an editor fact-checks and publishes.
How long does setup take, and what does AI usage cost?
About 30 minutes for the Space, connections, channels, and instructions, and four weeks to roll out all seven workflows. AI usage is metered at provider API rates against the workspace allowance, unless it runs through a connected Claude or Codex subscription.