Approach
Engineer what you control. Measure the rest honestly.
How we run engagements, the principles behind every recommendation, and the promises we refuse to make.
01 — Method
Discover → Diagnose → Engineer → Validate → Monitor
- 01
Discover
Understand the business, its audiences and domains, and which discovery channels actually drive outcomes for it. Agree what success means before touching anything.
Output: goals, domain list, channel priorities, access plan
- 02
Diagnose
Audit infrastructure, crawlability, content representation, agent readiness and conversion paths. Record a measurement baseline.
Output: prioritized findings register + baseline snapshot
- 03
Engineer
Implement only approved changes. Every change has an owner, a date, a rollback plan and a verification step.
Output: change log, pull requests, configuration records
- 04
Validate
Confirm each change works — status codes, rendered HTML, schema validation, crawler access, agent tests — then compare observable outcomes against the baseline.
Output: validation report with evidence
- 05
Monitor
Track visibility, traffic quality, regressions and conversions, with every discovery channel measured separately.
Output: alerts + monthly report
02 — Principles
What we believe about the AI-mediated web
Traditional SEO isn't obsolete.
Google says its AI features are rooted in its core Search systems, and that there are no extra requirements for them beyond being indexed and eligible for a snippet. Foundations come first.
There's no single "AI index".
OpenAI, Anthropic, Perplexity, Google and Microsoft each run different crawlers for training, search indexing and user-initiated fetches — with different controls. Policy has to be set per bot.
Access isn't recommendation.
A crawler fetching a page proves access. It doesn't prove retrieval, accurate representation, citation, referral or a sale. These are separate phenomena and we measure them separately.
Accessibility is agent-readiness.
Browser agents use the DOM and accessibility tree. Labeled forms, semantic structure and stable layouts help people with assistive technology and software agents alike.
Experimental means labeled.
We implement emerging conventions like llms.txt only with a stated purpose, known limits and an owner — and never sell them as ranking or citation levers.
Clients keep control.
Domains, registrars, analytics and code stay in your accounts. Our access is scoped, logged and revocable.
03 — Non-promises
What we will never promise
- Inclusion, ranking or citation in ChatGPT, Gemini, Perplexity, Copilot, Claude, Google AI Overviews, AI Mode or any other search or AI system.
- Specific ranking positions, traffic multipliers or timelines.
- That AI crawler activity means you're being recommended.
- That autonomous agents will adopt your service because it's discoverable.
- Results based on customers, testimonials or case studies we don't have.
Google's own guidance warns that no third-party tool has access to its internal ranking or AI systems. Any provider claiming otherwise deserves skepticism — including us.
04 — Reference implementation
This site is our first client.
aisideway.com is built the way we'd build yours:
- Static HTML generated at build time — every page is readable without JavaScript
- One canonical host, trailing-slash URLs, an accurate XML sitemap and a plain robots policy
- Organization, Service, Article and Breadcrumb structured data that matches visible content
- Self-hosted fonts, no third-party trackers, minimal JavaScript
- An llms.txt file, labeled experimental, with its purpose and limits documented in the repository
- A contact form protected by server-side Turnstile verification and a honeypot
Start with an assessment.
A fixed-scope review of your domains, search infrastructure, AI discoverability and agent readiness — with prioritized findings and a plan you can act on.