SEO tooling in the AI era: the system behind these pages

The tools that run modern SEO aren’t secret, and you can buy every one of them this afternoon. What’s changed in the AI era is not the tools but what you can do with them at once, because a single model can now hold the data from all of them in its head, read across the lot, and draft from the whole picture rather than one slice of it. This page is the working example, since every article on it was researched, scored and built with the same system we point at client sites, and the number under the proof is the same one we publish everywhere: 9,565 Bristol sites scored. Average: 42/100.

Last reviewed: 16 June 2026 · By Aaron Kennedy, founder of Originate. I built the pipeline this page describes, after three years and more than thirty website builds.

Contents

• Everyone has the tools
• The combination is the thing
• What each part actually does
• Why this was hard until recently
• This page is the proof
• What it means if you hire us
• Common questions

Everyone has the tools

Any agency can subscribe to the same software we use, so the stack on its own proves nothing. Ahrefs has been the backlink and keyword standard for years, DataForSEO sells the same search and SERP data to anyone with an API key, Perplexity is a consumer app, and Google gives away GA4 and Search Console to every site owner who asks. None of that is a moat, and an agency that lists its logos as if they were achievements is telling you it hasn’t found anything better to say. The tools are table stakes, the way a van and a ladder are table stakes for a window cleaner, and nobody hires the cleaner for the ladder.

So the honest question isn’t which tools we have, since the answer is the same boring list everyone reads off. The question is what we do with them that the next agency doesn’t, and that turns out to be the part almost nobody has built.

The combination is the thing

The thing worth claiming is the join, not the parts, because the difference is what happens when one model holds all of it at once. We pull keyword and authority data from Ahrefs, search-volume and live SERP results from DataForSEO, real visibility checks from Perplexity, behaviour from GA4 and query data from Search Console, then pipe the whole set into Claude as context for a single reading. The model sees the keyword that’s winnable, the page that’s already half-ranking for it, the query people actually type, and what the AI engines say about you today, all in one frame, and it reasons across them instead of one report at a time.

That join is the bit that didn’t exist a couple of years ago, and it’s the bit that’s hard to copy, because anyone can open five dashboards but very few have wired them into one context a model can think with. The output isn’t a prettier report. It’s a scored, sourced, drafted recommendation that already knows your numbers.

What each part actually does

Each tool covers a blind spot the others can’t see, which is the whole reason the combination beats any one of them. Ahrefs and DataForSEO between them tell us what’s worth chasing and how hard each term is, so we aim at the winnable ones rather than the flattering ones. GA4 and Search Console tell us what’s already happening on your own site, which is the difference between guessing and reading. Perplexity tells us whether the AI engines can see you at all, which is the question most SEO tools were never built to answer.

Then Claude reads across the four data sources at once and does the work a person would take a week to do, holding the keyword, the page, the query and the AI answer in one view. That’s the step that turns five subscriptions into one system, and it’s why the recommendation that comes out the other end is specific to your site rather than generic to your sector.

Why this was hard until recently

This setup wasn’t possible at any sane cost until models got big enough to hold real context, which is a genuinely recent change. A couple of years ago you could pull all the same data, but you’d be the one reading it, cross-referencing it, and turning it into a plan, and that’s days of expensive analyst time per site that only an agency with a procurement budget could absorb. The data was available, the joining of it was not, because the joining was a human bottleneck.

What moved is the cost of the reading, not the cost of the data. A model that can take the full picture as context collapses the analyst week into a single pass, which is what lets a three-person studio run the kind of pipeline that used to need a team. That’s the shift, and it’s why the line we use internally is that the data was always there, the affordable way to think with all of it at once is the new part.

This page is the proof

The strongest proof that this works is the page you’re reading, because we ran the system on ourselves before we ever pointed it at a client. Every article in this section was researched with that pipeline: the keyword pulled and scored for difficulty, the live SERP read for what already ranks, the AI engines checked for what they say about the topic, and the draft built from all of it against a strict house style. We also ran it on our own visibility and published the result, including the parts where we scored zero, because a method you only show when it flatters you isn’t a method, it’s a brochure.

The Bristol leaderboard is the same story at city scale, where we scored 9,565 Bristol sites against the five pillars and the average came back at 42 out of 100, with only 6% passing Google’s speed benchmarks and the typical site taking 10 seconds to load on a phone. Every one of those scores sits in public with the business name next to it, which is proof no slide deck can fake.

What it means if you hire us

For a client, the system is the reason the top Foundations tier can promise something most agencies can only gesture at, which is SEO and AI discoverability that actually compounds rather than just bills. The pipeline runs on your site monthly, finds the winnable terms, checks whether the AI engines can see you, drafts the page that closes the gap, and shows its working every step, so you can watch the method instead of trusting it. You get the output of an analyst team at the cost of a subscription, which is the whole economic point of building it this way.

You don’t have to take any of that on faith, which is rather the point of writing it down here. Run the Diagnostic on your own site and you’ll see the front end of the same system, the score and the actual fixes, before anyone asks you for a penny. If the result is good, you can ignore everything I wrote. If it isn’t, at least now you know exactly what you’ve been avoiding, and which part of it is worth spending money on first.

Common questions

What SEO tools does Originate use? We use Ahrefs for keyword and authority data, DataForSEO for search volume and live SERP results, Perplexity for AI-visibility checks, and Google’s own GA4 and Search Console for behaviour and query data. The tools are standard, and we say so. What’s different is that we pipe all of them into one model, Claude, for context, so the analysis reads across every source at once rather than one report at a time.

Isn’t that just the same tools everyone has? The tools are the same, the join is not. Most agencies open each dashboard separately and a person stitches the findings together by hand, which is slow and expensive enough that it rarely happens properly. We wire the five sources into a single context a model can reason with, so the keyword, the page, the query and the AI answer get read together. That combination is the part that’s genuinely hard to copy.

Why does AI make a difference to SEO tooling? The data was always available, but holding all of it in your head at once was a human bottleneck that cost days of analyst time per site. A model that can take the full picture as context collapses that week into a single pass, which is what lets a small studio run a pipeline that used to need a whole team. The change is the cost of the reading, not the cost of the data.

Can I see the system working before I pay? Yes, and that’s deliberate. This whole Writing section was built with the pipeline, the Bristol leaderboard publishes 9,565 real scores from it, and the Diagnostic gives you the front end of the same system on your own site for free. Run it, read your score, see the fixes, and decide from there.

Does the AI write the articles for you? The system does the research, the scoring and a first draft, then a strict house style and a human pass decide what actually ships, because an unedited model draft reads like an unedited model draft. The point isn’t to remove the writing, it’s to remove the week of reading that used to come before it.

---

Opinions are cheap until you put a number against your own site. The Diagnostic is free and already run on 9,565 Bristol sites, so look yourself up in seconds, or ask for a fresh one and have your score in a couple of days. The score comes first.

Run the Diagnostic · See Foundations

Common questions

What SEO tools does Originate use?

We use Ahrefs for keyword and authority data, DataForSEO for search volume and live SERP results, Perplexity for AI-visibility checks, and Google's own GA4 and Search Console for behaviour and query data. The tools are standard, and we say so. What's different is that we pipe all of them into one model, Claude, for context, so the analysis reads across every source at once rather than one report at a time.

Isn't that just the same tools everyone has?

The tools are the same, the join is not. Most agencies open each dashboard separately and a person stitches the findings together by hand, which is slow and expensive enough that it rarely happens properly. We wire the five sources into a single context a model can reason with, so the keyword, the page, the query and the AI answer get read together. That combination is the part that's genuinely hard to copy.

Why does AI make a difference to SEO tooling?

The data was always available, but holding all of it in your head at once was a human bottleneck that cost days of analyst time per site. A model that can take the full picture as context collapses that week into a single pass, which is what lets a small studio run a pipeline that used to need a whole team. The change is the cost of the reading, not the cost of the data.

Can I see the system working before I pay?

Yes, and that's deliberate. This whole Writing section was built with the pipeline, the Bristol leaderboard publishes 9,565 real scores from it, and the Diagnostic gives you the front end of the same system on your own site for free. Run it, read your score, see the fixes, and decide from there.

Does the AI write the articles for you?

The system does the research, the scoring and a first draft, then a strict house style and a human pass decide what actually ships, because an unedited model draft reads like an unedited model draft. The point isn't to remove the writing, it's to remove the week of reading that used to come before it.