- One marketer has rebuilt most of her SEO content workflow around Claude Fable 5, using it to coordinate research, briefs, drafting, QA and publishing.
- The setup still depends on live SEO data and human editorial judgment, suggesting AI is not replacing the whole SEO stack but is starting to replace the workflow built around it.
AI is starting to unbundle the SEO software stack.
The data layer still matters. Keyword volumes, live SERPs, backlink indexes and crawl data do not disappear. But the software sitting between that data and the finished article is becoming easier to rebuild with AI.
One recent workflow makes that shift unusually clear.
how I run my SEO vertical with Fable 5, instead of paying for an SEO tool
Fable 5 is very capable, it plans the entire funnel, then hands the build to sonnet, opus and GPT sub agents
an SEO tool largely sells you keyword data, SERP scraping, and content gap analysis, then hands… pic.twitter.com/rElvXNs4Ij
— Shann³ (@shannholmberg) July 8, 2026
The thread describes a system where Claude Fable 5 acts as the planner, then hands execution to cheaper models such as Sonnet, Opus and GPT-based sub-agents.
The workflow covers research, content gaps, briefing, drafting, visuals, QA and publishing. In other words, it reproduces much of the production layer that SEO platforms increasingly sell as an add-on to their core data.
The interesting part is not that AI can write an article.
It is that AI can now coordinate the machinery around it.
What the Workflow Actually Replaces
The setup starts with a topic or keyword seed.
From there, Fable 5 pulls live search data, reviews the current SERP, extracts competitor pages and analyzes intent, structure and content gaps. That research then feeds into a brief, an outline, a draft, visual planning and a QA pass before the article is prepared for publication.
The process is not especially exotic. Most modern SEO content platforms already offer some version of the same sequence:
- keyword enrichment;
- SERP analysis;
- competitor research;
- content gap detection;
- brief generation;
- drafting;
- optimization and QA.
The difference is that those steps are no longer tied to one vendor interface.
An agentic workflow can connect the data source directly to the team’s own prompts, rules and publishing process. That gives marketers more control over what the system values and how the final output is structured.
The workflow can be adjusted around editorial priorities instead of accepting a fixed content score or generic brief template.
That is where the pressure on SEO tools starts.
The more easily teams can recreate the analysis and production layer themselves, the harder it becomes for software companies to defend features that mainly repackage the same underlying data.
AI Still Needs Real SEO Data
The workflow does not run on model memory alone.
It still needs live data from providers such as DataForSEO, Ahrefs or another source capable of supplying current SERPs, keyword metrics and competitor information.
Without that feed, the model is guessing.
This is why the claim that Fable 5 replaces an SEO tool is only partly true. It may replace the interface, the brief generator and the writing layer. It does not replace the infrastructure collecting search data.
That distinction matters because the data is the hardest part of the stack to reproduce.
Keyword databases require constant crawling, processing and storage. Backlink indexes require enormous infrastructure. Rank tracking needs fresh queries at scale. These are not tasks a language model performs by itself.
The likely outcome is not that teams stop paying for SEO data.
It is that they become less willing to pay for every workflow layered on top of it.
A marketer may still keep Ahrefs or Semrush for data, but drop a separate content brief platform, AI writer, optimization score and QA tool if one agentic workflow can handle those steps.
The stack does not disappear.
It gets thinner.
The Editorial Angle Is Still the Hard Part
The strongest part of the workflow is also the part that remains human.
The thread makes clear that the system can collect research, compare competitors and surface gaps, but the editorial angle stays with the person running the process.
That is an important limit.
AI can tell a team what already ranks. It can identify repeated headings, missing subtopics and common claims. It can summarize the market quickly.
It is much less reliable at deciding what the article should actually say that is new.
That decision requires context, taste and judgment.
A page can be perfectly optimized and still have no reason to exist. It can cover every keyword variation and still add nothing. It can mirror the SERP so closely that it becomes indistinguishable from the content already there.
This is why the brief matters more than the draft.
The model can build from a strong editorial direction. It cannot reliably invent one on command.
For publishers and content teams, that separation should stay explicit: AI handles the research load, but humans decide the story.
One Model Plans, Cheaper Models Execute
The workflow also shows how frontier models are increasingly being used.
Fable 5 is not asked to perform every task. It acts as the planner and reviewer, while cheaper models handle the higher-volume work.
That model hierarchy is becoming common in agentic systems.
The strongest model keeps the project on track, decides what happens next and checks whether the output meets the goal. Less expensive models handle drafting, formatting, summarizing or repetitive sub-tasks.
Anthropic positioned Fable 5 around long-horizon agentic work, including multi-stage planning, delegation and extended workflows. That makes it better suited to coordinating a content pipeline than to producing one isolated answer.
Cost is part of the design.
Search Engine Journal reported API pricing of $10 per million input tokens and $50 per million output tokens. Running every step through the most expensive model would waste much of the economic advantage.
Using Fable 5 as the architect and cheaper models as the workforce is what makes the setup viable.
This reflects a broader change we covered in our analysis of Google’s agentic web push: models are moving from answering questions to operating workflows.
The Skill Layer Becomes the Real Moat
The model alone is not enough.
The workflow also depends on reusable skill files that define how the agents research, write, review and publish.
These files can carry the publication’s voice, structural rules, sourcing standards, internal linking requirements, preferred paragraph length and common failure modes.
The thread calls this a voice DNA or anti-slop layer.
Without it, even strong research can produce generic content.
That is the part many AI content setups underestimate. Teams compare models endlessly, but spend less time building the instructions that make the model useful inside a real editorial process.
A blank chat starts from zero every time.
A skill-based system does not.
It carries forward the rules the team has already learned. The model can improve, but the accumulated workflow knowledge stays with the business.
That may become the more defensible advantage.
Everyone can access the same frontier model. Not everyone has the same editorial rules, QA logic, source standards or publishing system around it.
Why This Matters for SEO Software
The pressure is highest on tools whose main value sits between data and execution.
Features such as automated briefs, outline generation, keyword coverage scores, content gap summaries and AI drafts are becoming easier to reproduce.
They are useful, but they are no longer especially difficult to build.
The more capable models become at coordinating multi-step work, the more those features look like workflows rather than standalone products.
That does not mean established SEO platforms are in immediate trouble.
Their proprietary data remains valuable. Their integrations, support and simplicity still matter. Many teams will prefer a stable interface over maintaining APIs, model routing and custom skills.
But the value hierarchy is changing.
The data layer becomes more defensible.
The generic workflow layer becomes less so.
As we argued in our analysis of AI search exposing weak SEO systems, automation is most useful when it handles repeatable operational work without pretending to replace strategy.
That is exactly what this setup does.
What SEO Teams Should Take From This
The practical lesson is not to cancel every SEO subscription.
It is to audit what each tool actually does.
Which platforms provide data your team cannot reproduce? Which ones mainly turn that data into briefs or drafts? Which interfaces save real time? Which steps could already be handled by an internal agent using your own rules?
For some teams, the answer will still be a full SEO platform.
For others, the stack may shrink to a data provider, a crawler, a rank tracker, a CMS and an agentic layer connecting the rest.
The trade-off is clear.
A ready-made platform offers convenience and lower maintenance. A custom workflow offers flexibility, control and the ability to shape the process around the team’s own editorial standards.
Neither approach is automatically better.
But the fact that the choice now exists is the real story.
The Takeaway
Claude Fable 5 is not replacing SEO tools outright.
It is starting to replace the workflow wrapped around SEO data.
Keyword databases, live SERPs, crawl information and human editorial judgment still matter. But the work connecting research to a finished article is becoming programmable.
That is a bigger shift than AI writing.
The writing layer was always easy to copy.
The workflow around it is now becoming easy to rebuild too.
