- A Favikon-ranked local SEO consultant is sharing a Claude setup that treats business context, competitor data and buyer intent as the real starting point for AI-assisted SEO.
- The workflow points to a wider shift: useful AI SEO output increasingly depends less on one clever prompt and more on the context system built before the task begins.
Most weak AI SEO output has the same problem.
It sounds clean. It is structured. It may even include a checklist. But it could apply to almost any business in almost any market.
That is the weakness Sarvesh Shrivastava, founder of Alventra Marketing and a Favikon-ranked local SEO consultant, is trying to fix in his Claude workflow. His argument is simple: bad AI SEO output is often not a model problem. It is a context problem.
Just my personal opinion here:
Being in real estate and not leveraging SEO or AI Search Optimization in 2026 is like running ads to a business with no landing page.
Here is an exact step by step guide to fix that.
Let’s start with the real issue.
Here is where buyers now look… https://t.co/YtxS3RXM2w pic.twitter.com/tu1Oax1uJn
— Alex Groberman (@alexgroberman) June 26, 2026
In the X thread, Shrivastava outlines a setup he runs before asking Claude for SEO work. The interesting part is not that he uses AI. It is what he loads before the first prompt.
The workflow is built around business memory, competitor evidence, model selection, standing instructions and keyword intent filters. That makes it more useful than another viral prompt list.
It shows where AI-assisted SEO is moving: away from one-off prompting and toward repeatable operating systems for specific businesses, markets and search intent.
The Workflow Starts Before the Prompt
The first step is loading what Shrivastava calls the business brain.
That means Claude is given the business name, website, location, services, target cities and closest competitors before the SEO task begins. The goal is to stop the model from answering as if every local business has the same priorities.
That matters because local SEO is not generic by design.
A dentist in Austin, a roofer in Chicago and a personal injury lawyer in Miami may all need better search visibility, but they do not need the same service pages, keyword targets, review strategy, local content or Google Business Profile priorities.
Without that context, the model fills the gaps with broad advice: optimize your GBP, build local citations, create helpful content, improve on-page SEO. None of that is wrong. It is just too shallow to be useful.
The better first prompt is not “write me an SEO strategy.”
It is: here is the business, here is the market, here are the services, here are the competitors, here is what we are trying to win.
What the Context Layer Should Include
The most useful part of the workflow is the structure behind it.
For a local SEO client, the context layer should usually include:
- business name, website and primary location
- main services and service categories
- target cities, neighborhoods or service areas
- ideal customer type and conversion goal
- Google Business Profile URL and current GBP categories
- top competitors with website and GBP URLs
- review count, average rating and obvious review gaps
- priority services that generate real revenue
- services or keywords the business does not want to target
- existing pages that should be improved before new pages are created
That list does more than help Claude write better copy.
It gives the model boundaries. It tells the system what market it is working inside, which services matter commercially and which competitors should shape the recommendations.
That is why context loading is more important than prompt polish. A perfect prompt with no market evidence still produces generic strategy. A simple prompt with strong business context can produce something much closer to usable work.
Claude Cowork Changes the Shape of the Task
Shrivastava also points to Claude Cowork, Anthropic’s agentic AI system for knowledge work, rather than treating Claude as a standard chat window.
That distinction matters for SEO because many SEO tasks are not single answers.
A useful local SEO workflow may involve reading a competitor file, comparing GBP categories, grouping service keywords, checking city-page gaps, drafting page briefs and turning scattered information into a practical plan. That is closer to a working loop than a normal chat response.
Shrivastava also names Claude Opus 4.7 and adaptive thinking as part of the setup. The important point is not that one model setting magically solves SEO. It is that model choice, interface and task structure now affect the quality of the work.
For marketers, that is a shift.
The tool stack is no longer separate from the SEO process. It is part of the process.
The Competitor File Is the Practical Layer
The competitor file may be the most valuable part of the setup.
Shrivastava describes a client-specific file that includes competitor websites, Google Business Profile URLs, review counts, ratings, ranking gaps and missing GBP categories.
That is exactly the kind of evidence AI needs before it can stop producing surface-level local SEO advice.
Without competitor data, Claude can only say what a business should usually do. With competitor data, it can compare the client against the market in front of it.
That changes the questions:
- Which services do ranking competitors make more visible?
- Which GBP categories appear across the strongest competitors?
- Which city or neighborhood pages does the client lack?
- Where is the review gap large enough to affect trust?
- Which keyword gaps are tied to people ready to call?
This is where AI becomes useful for local SEO.
Not because it replaces an SEO professional, but because it can organize messy inputs and expose gaps faster once the right evidence is loaded.
Keyword Intent Comes Before Keyword Expansion
The keyword filter is another important part of the workflow.
For local SEO, Shrivastava narrows Claude toward buyer-intent patterns: service plus city, emergency plus service, “near me” variations and phrases that suggest someone is ready to call, book or request a quote.
That matters because AI tools are good at expanding keyword lists, but weak at judging commercial value unless the user gives them rules.
A plumber does not need 200 blog topics before the emergency repair page is useful. A roofer does not need a long informational calendar if the roof replacement and storm damage pages are thin. A dentist does not need generic oral health articles before the core service pages and map-pack signals are in place.
Google’s own local ranking documentation still frames local results around relevance, distance and prominence. That makes vague keyword expansion less useful than matching the right services, locations, proof signals and buyer intent.
This is also why local SEO professionals are watching high-intent map-pack visibility closely. The search journey is not only about ranking an article. It is about being visible when someone is ready to act.
The Checklist Turns AI Into a Process
The final layer is a pre-session checklist.
Before starting, Shrivastava checks whether the work is happening in Cowork, whether the selected model and thinking mode are active, whether the business context is loaded, whether the competitor file is available and whether the keyword intent filter has been set.
That sounds mechanical, but it is what makes the workflow repeatable.
AI SEO output often fails because teams skip setup and then blame the model. A checklist forces the user to confirm the conditions that make a useful answer possible.
For agencies, that matters even more.
A single strategist using Claude well is helpful. A team using the same business brief, competitor format, intent filter and review process can turn AI from a personal shortcut into a workflow standard.
What Marketers Should Take From This
The thread points to a broader change in AI-assisted SEO.
The question is no longer only “What prompt should I use?”
The better question is: what does the model need to know before the prompt starts?
For marketers, that changes the workflow:
- Load business context before task instructions.
- Use competitor files instead of vague competitor descriptions.
- Filter keyword research by buyer intent before expanding lists.
- Use standing instructions to keep the strategy frame consistent.
- Match model and interface choices to the complexity of the task.
- Check AI output against real market evidence before using it.
This does not mean Claude, or any other AI system, can run local SEO without expert review.
It still needs accurate inputs. It still needs human judgment. It can still miss local nuance, over-prioritize recent context or produce recommendations that sound better than they perform.
But the workflow explains why some AI SEO output feels useful while most of it feels disposable.
The difference is rarely one better prompt. It is the operating context around the prompt.
That gap matters more as AI SEO advice floods the market while many brands still miss the basics. A business that has not defined its services, markets, competitors and buyer intent will not fix the problem by asking a stronger model for a better answer.
The Takeaway
Claude workflows like Shrivastava’s show where practical AI SEO is heading.
The useful work is moving away from clever prompts and toward structured context systems: business briefs, competitor files, standing instructions, intent filters and repeatable checks before the model starts producing output.
That is less exciting than another viral prompt.
It is also closer to how professional SEO actually works.
Models can generate drafts, plans and recommendations quickly. But for local SEO, speed is not the bottleneck. Specificity is. The businesses getting value from AI will be the ones that load the market reality first and make the model work inside it.
