That view is your dev sprint request, ready to paste, with an argument already attached to each line.
``` `Verified date` is the column everyone leaves out, and it is the one that catches things. Filter for `Status = Fixed AND Verified date is empty`. On every account I have run this on, that view has had rows in it — things marked done by a developer that never actually shipped, or shipped and regressed. Check it monthly. ## System 5: the AI visibility log There is no mature tooling for this yet, which is exactly why a Notion database is genuinely competitive here rather than a compromise. | Property | Type | | --- | --- | | Prompt | Title | | Platform | Select — ChatGPT / Google AI Overview / Perplexity / Claude / Gemini / Copilot | | Date checked | Date | | **Brand mentioned?** | Checkbox | | **Cited as source?** | Checkbox | | Competitors mentioned | Multi-select | | Answer snippet | Text | | Our page that should have won | Relation → Content | | Gap identified | Text | Check 20–30 priority prompts monthly. Group by Platform and you have a month-over-month trend nobody else on the account is producing. They are different outcomes. **Mentioned** is brand exposure inside an answer. **Cited** is a linked source — a click, and a trust signal. Collapsing them into one "AI visibility" number is the most common measurement mistake in this area, and it is why two agencies quoting the same study reach opposite conclusions. I have written up the methodology problem in [the AI search visibility benchmark](https://thejayant.in/blog/ai-search-visibility-benchmark), and what actually drives source selection in [how AI systems choose their sources](https://thejayant.in/blog/how-ai-chooses-sources). The `Gap identified` column is what makes this a system rather than a report. If Perplexity cites three competitors on a prompt and none of them answer it well, that is a brief — and because the log relates to Content, the brief lands in the same pipeline. Pair it with an **Answer Bank**: one row per question customers actually ask, with a 40–60 word direct answer written in extractable form. Those answers get placed under matching H2s across the site. That is the practical core of answer engine optimisation — writing quotable units, then tracking whether they get quoted. Tactics are in [how to get cited by AI search](https://thejayant.in/blog/how-to-get-cited-by-ai-search). ## System 6: the backlink and digital PR CRM Notion beats most dedicated outreach CRMs for one reason: the pitch, the notes and the relationship history attach to the same record, and that record relates to the page you are building links to. Properties: Site · DR/DA · Contact name · Email · Status (Prospect → Contacted → Replied → Negotiating → Live → Dead) · Last touch · **Next touch** (formula) · Link target (relation → Content) · Anchor · Cost · Notes. ``` dateAdd(prop("Last touch"), 7, "days")One automation — when Status → Contacted, set Last touch = today — plus one view filtered to Next touch ≤ today AND Status not in (Live, Dead), and your daily follow-up list builds itself from a single date rule. No reminders, no sequencing tool.
``` ## System 7: the client reporting hub One Reporting database, one row per client per month, created automatically on the 1st by a recurring automation. Each row's page uses a template with fixed sections: Executive summary · Traffic · Rankings · Conversions · What we did · What we found · Next month. Data arrives one of three ways, in increasing order of effort and decreasing order of ongoing pain: paste it, pull it with Zapier or Make, or write a Notion Worker that hits the GA4 and Search Console APIs on the 1st and writes straight into the page. Then a Custom Agent scoped to that database drafts the executive summary from numbers already on the page. Reviewing a draft takes about four minutes; writing one from scratch takes twenty-five. Across eight clients that is roughly three hours back every month. An agent drafting a client summary will produce confident sentences about things it half-knows. Three categories to verify every time: **Every number.** Agents transpose and round. If it says traffic rose 34%, check that it rose 34%. **Every causal claim.** "Traffic rose because of the content refresh" is a hypothesis the agent has no way to test. Either substantiate it or soften it. **Anything that sounds impressive.** Confident phrasing is the model's default register, not evidence. If a sentence would embarrass you in a QBR, cut it. ## System 8: the SOP library Every recurring task written once: new client onboarding, monthly audit, content QA, migration checklist, schema implementation, Search Console troubleshooting. Structure: Process name · Category · Owner · Last reviewed · Steps (numbered, with screenshots) · Common failure modes · Tools needed. Two payoffs, and the second is the one people do not expect. First, delegation stops being a conversation. Second — **this becomes the ground truth for every AI agent you build.** An agent pointed at a well-written SOP database produces dramatically better output than one guessing at how your team works. If you only build one of these nine systems before adding AI, build this one. ## System 9: competitor watch Properties: Competitor · Domain · Tracked pages (relation) · Last content published · Content velocity per month · New pages spotted (rollup) · Their AI Overview presence · Notes. A Worker or a Zapier flow drops new URLs from their sitemap into a New Pages database weekly. You see what a competitor is publishing before it ranks, not after — which is the difference between responding and reacting. ## How the nine connect The architecture matters more than any individual database. Six relations do all the work: ``` Keywords ──────────► Content ◄────────── Backlinks │ ▲ │ └──── AI Answer Log │ ├──► Content (internal links, two-way) │ ├──► Technical Audit (URLs affected) │ └──► Reporting (pages covered this month) SOPs ──────► (context for every agent)Read it as a hub and spokes. Content is the hub. Keywords say what a page should win, backlinks say what points at it, the AI Answer Log says whether models cite it, the audit tracker says whether it is technically sound, and reporting says what happened to it. One record, five perspectives.
``` The practical consequence: you can answer "what is the state of this page?" in one click, and "which pages are the weakest across all five dimensions?" with one sorted view. That question is essentially unanswerable in a spreadsheet stack. ## What Notion cannot replace Be honest with clients about this, because overselling it is how the system loses credibility in month two. - **A crawler.** Screaming Frog or Sitebulb finds the issues. Notion tracks them. - **A rank tracker.** Notion stores positions; it does not measure them. Something has to fetch the data. - **An analytics platform.** GA4 and Search Console remain the source of truth. Notion holds the narrative around the numbers. - **Log file analysis.** Volume alone rules it out. - **Scale beyond roughly 20,000 rows.** Filtering and rollups slow noticeably past that. Archive completed records into a separate database rather than letting one grow forever — a large keyword database on a big ecommerce account will hit this. Notion is the operating layer. It replaces the pile of spreadsheets, docs, Trello boards and Slack threads sitting _between_ those tools and a decision — which is where most of the hours actually go. ## What to build first Do not build nine systems. Build two, use them for a month, then add. 1. **Keywords + Content with the relation between them.** This is 80% of the value. The gap list and the priority order fall out of it immediately. 2. **The brief template**, because briefs are the bottleneck and the template is twenty minutes of work. 3. Then whichever of the remaining seven maps to your biggest time sink — usually reporting for agencies, the audit tracker for in-house. Agents come after all of it. An agent pointed at three weeks of consistent, structured data produces work you can ship; one pointed at an empty workspace produces confident nonsense. ## FAQ ### Is Notion good for SEO? Notion is very good as the operating layer for SEO work — keyword databases, content pipelines, briefs, audit trackers, AI visibility logs and client reporting all work better there than in spreadsheets, because relations and rollups keep everything connected as things change. It is not a replacement for a crawler, a rank tracker or an analytics platform, which remain the data layer. ### Can Notion replace a spreadsheet for keyword research? Yes, and it solves three things a spreadsheet cannot. Keywords relate directly to the content targeting them rather than through a VLOOKUP that breaks. Cannibalisation becomes visible through a formula that flags any keyword related to more than one page. And a filter for keywords with no target URL is a permanently current content gap list, rather than a quarterly exercise you rebuild from scratch. ### How do I find orphan pages in Notion without a crawler? Create a two-way relation from your Content database to itself, called something like "Internal links planned". When briefing each post, record which existing pages it will link to. The reverse side of that relation gives you an inbound internal link count per page automatically. Sort that rollup ascending and the pages at the top are your orphans — always current, no crawl needed. ### How do you track AI Overview and ChatGPT visibility in Notion? Build an AI Answer Log database with one row per prompt per platform, and record two separate checkboxes: whether your brand was mentioned, and whether it was cited as a source. Add competitors mentioned, the answer snippet, a relation to the page that should have won, and a gap column. Check 20–30 priority prompts monthly and group by platform for a month-over-month trend. Keeping mention and citation separate is essential — they are different outcomes. ### How many databases does an SEO Notion setup need? Nine covers a full agency workflow: Keywords, Content, Technical Audit, AI Answer Log, Backlinks, Reporting, SOPs, Competitors and New Pages. Most people should start with two — Keywords and Content, with a relation between them — which delivers roughly 80% of the value. Add the rest only as a specific time sink justifies each one. ### Can Notion AI write my content briefs? It can draft them well if two conditions are met: the brief is a page template with fixed headings, and the agent is scoped to your keyword database rather than the whole workspace. Trigger it when a content row moves to Brief status. You then edit rather than write, which saves roughly 40 minutes per brief. The fixed structure is what makes the output consistent enough to be useful. ### How big can a Notion SEO database get before it slows down? Comfortably into the low thousands of rows. Past roughly 20,000, filtering and rollups slow noticeably — which a large ecommerce keyword set will reach. The fix is archiving completed or irrelevant records into a separate database rather than letting one grow indefinitely. ### Is Notion better than Airtable for SEO work? It depends on the balance of writing and data in your process. Airtable is the stronger database — more field types, better at scale, more robust automations — so it wins when you are managing tens of thousands of records with complex logic. Notion wins when briefs, SOPs, reports and notes need to live alongside the structured data, which is true of most SEO work. Many teams use Notion as the hub and keep one large dataset elsewhere. ### What Notion plan do I need for an SEO workflow? Free is enough to build the databases, relations and rollups and prove the system works. Automations need a paid plan for the Slack action. The AI features — the personal Agent, Custom Agents for brief and report drafting, AI Meeting Notes — require Business at $20 per member per month, plus Notion credits for Custom Agent usage. If you are running client work, Enterprise adds zero data retention with LLM providers.