I keep thinking about this website as a strange little workshop: part portfolio, part notebook, part test site, part public proof of process. It is built with AI tools, measured with search and analytics tools, then edited by a person who is trying very hard not to let the robots sand off the useful edges.
That is what I mean by created by robots for robots. Not in the gimmicky sense. I do not want a site that sounds like a prompt escaped into the wild. I want a site that is clear enough for people to trust, structured enough for search engines to understand, and specific enough for AI systems to quote when they need a useful answer.
The Double Atari site has become a working example of the consulting process I use with clients. Build something clean. Put measurement in place. Watch the data. Let the weak signals point toward better content, better links, better structure, and better answers.
What does created by robots for robots mean?
It means the site is shaped by a practical tension. I use AI to help build, audit, organize, and improve the site, but the finished page still has to feel like it came from a person with a point of view.
The robotic part is useful. AI can help produce HTML, scan a page for schema issues, compare metadata patterns, group keywords, propose internal links, and notice when a post is answering a question indirectly instead of directly. Search crawlers, AI assistants, and answer engines also read the site in a robotic way. They look for entities, headings, page relationships, schema, author signals, and concise answers.
The human part is the filter. I decide what is worth saying, what should be cut, what sounds false, where the useful example belongs, and which tradeoffs are acceptable. A robot can suggest ten headings. It cannot know whether the page sounds like something I would actually say to a founder, marketer, or internal team trying to make sense of SEO and AI search.
That balance matters because a site built only for machines gets weird quickly. A site built only for human readers can also miss the structure that helps it get discovered. The goal is to make both audiences comfortable without pretending they are the same audience.
How did the Double Atari site become an AI-assisted build?
The first version started after I saw a friend's AI-built Scrabble site and wrote about it in the post that kicked off the Double Atari rebuild. The site was fast, simple, and much better than I expected from someone who was not usually building websites. That made me curious.
I already knew the old Double Atari site needed work. It did not reflect the SEO, GEO, AEO, analytics, CMS, and content strategy work I was doing. It also did not give me enough freedom to test things quickly. So I used AI as a building partner, not just a writing partner. It helped with static HTML structure, navigation patterns, schema, blog templates, service page consistency, and quality checks.
That first pass was not magic. It was still a lot of reviewing, rewriting, fixing, and checking. But it changed the speed of iteration. Instead of waiting for a large rebuild cycle, I could make the site better in smaller passes. A title tag here. A clearer service page there. A new FAQ block, a better author card, a cleaner internal link, a sitemap update.
The site is now less of a finished artifact and more of a living system. I wrote about that in How the Double Atari Website Keeps Evolving, and this post is the next layer: how the AI build process connects to measurement, discovery, and the reality that authority is still the hard part.
How I use Perplexity with GA4, GSC, and SEMrush
Perplexity is most useful when it is grounded in data instead of asked to invent a strategy from scratch. I use it as a second brain for interpreting exports and turning them into a work plan.
Google Search Console gives me the starting point. I look for non-branded queries, pages with impressions but weak clicks, posts sitting in positions 15 through 45, question queries that deserve clearer answers, and pages that Google seems to be testing but not fully rewarding yet. Those are not failures. They are clues.
GA4 tells me whether the visits that do arrive are worth more attention. If a page has low traffic but strong engagement time, good scroll depth, or contact activity, I do not ignore it just because the session count is small. That matters more now because AI search and answer surfaces may reduce click volume while increasing the value of the clicks that remain.
SEMrush helps me check the outside world. I use it to review keyword gaps, position movement, SERP features, competitor pages, and backlink patterns. If Search Console shows a near-miss query and SEMrush shows competitors winning with deeper definitions, better examples, or stronger supporting pages, that usually tells me what kind of edit to make.
Then Perplexity helps pull those signals together. I can ask it to group queries into clusters, identify which service page or blog post should be the target, suggest internal links from existing articles, review whether the page answers the query directly, and build a short backlog of edits. The key is that the data comes first. The AI helps with synthesis, not fantasy.
What does the site improvement loop look like?
The loop is simple enough that I can actually keep doing it.
- Collect the signals. Pull Search Console queries, GA4 engagement data, SEMrush keyword movement, backlink changes, and any manual AI search citation checks.
- Choose the page, not just the keyword. Decide whether the opportunity belongs on an existing service page, an existing post, a new post, or a stronger internal link path.
- Audit the current page. Check title tag, meta description, H1, intro, headings, examples, internal links, schema, author signals, and whether the answer appears early enough.
- Edit for usefulness first. Add the missing explanation, practical steps, examples, questions, comparison language, or measurement notes before worrying about polish.
- Validate the robotic layer. Confirm clean URLs, canonical tags, Article schema, BreadcrumbList schema, FAQPage schema where appropriate, sitemap entries, and internal links.
- Measure again. Watch impressions, clicks, average position, engagement, inquiries, and AI search mentions over time.
This is not glamorous work, but it is the work that compounds. The site gets clearer each time the loop runs.
What kinds of updates have made the site better?
The most useful updates have been boring in the best way.
I have tightened page titles so they match the actual search intent. I have rewritten meta descriptions so they sound less like placeholders and more like reasons to click. I have added FAQ sections where questions deserve direct answers. I have used schema markup to make the page type, author, breadcrumb, and common questions easier to interpret.
I have also been improving internal links. A post about answer engine optimization should point naturally to the GEO and AEO service page. A post about analytics should point to analytics and reporting. A post about keyword maps should connect to content strategy. Those links help readers move through the site, and they help machines understand which pages are central.
The other improvement is voice. The more I use the site, the more I want it to sound like a practitioner explaining what he is actually doing. I would rather publish a smaller post with a real process than a polished but generic article that could live on any agency blog.
Where is authority still hard?
Authority is the honest problem.
It is easier to increase content volume than it is to increase trust. It is easier to add schema than it is to earn a credible mention. It is easier to publish another post than it is to get a useful backlink from a relevant site, a local organization, a podcast, a newsletter, or a partner resource.
That is where the Double Atari site is still developing. The content and optimization work have moved faster than the authority signals. I can see the site becoming more complete, but authority has not grown as quickly as the page count, internal linking, or structured data.
That does not make the content work wasted. It means the next phase has to include reputation-building. Better resources. Better outreach. More useful examples. More reasons for other people to reference the site. Search engines and AI systems both need evidence that the site is not just well organized, but worth trusting.
What do I want to document next?
I do not want to pretend the site has already won. I want to document whether the work is moving the right needles. These are the measurements I care about next.
- Non-branded impressions and clicks. Track month-over-month growth in Search Console for queries that do not include Double Atari or my name.
- Near-miss rankings. Watch terms currently sitting around positions 15 through 45 and document which ones move into the top 10 after targeted updates.
- Referring domains and quality backlinks. Track not just the count of referring domains, but whether the links come from relevant, credible sites.
- AI search citations and mentions. Run a fixed set of prompts in tools like Perplexity and ChatGPT, then document whether Double Atari appears as a cited or mentioned resource.
- Consultation inquiries and engagement. Connect content work to contact form activity, engaged sessions, scroll depth, and time on page instead of judging every post by traffic alone.
Those goals are intentionally measurable, but they are not fake achievements. They are the scorecard I want to keep in public view as the site matures.
How does this change how I think about SEO, GEO, and AEO?
It makes the old boundaries feel less useful.
SEO, GEO, and AEO still have different meanings, but the practical work overlaps. A strong page needs to rank, answer, and be citeable. It needs clear structure for traditional search, concise explanations for answer engines, and enough credibility for generative systems to feel safe referencing it.
That is why I keep coming back to the same fundamentals: clean technical structure, useful content, specific examples, schema, internal links, author signals, measurement, and authority. The tools keep changing, but the discipline is still about helping the right reader find the right answer and trust the person who wrote it.
FAQ
Is using AI to build a website bad for SEO?
No. The problem is not using AI. The problem is publishing generic, unreviewed, unhelpful pages. AI can help with structure, drafts, schema, audits, and quality checks, but a human still needs to make the page accurate, useful, and specific.
How does Perplexity help with SEO and website updates?
Perplexity is useful for synthesizing GA4, Google Search Console, SEMrush, crawl exports, and page reviews into a practical backlog. It helps connect signals across tools, but the strongest recommendations come from real performance data.
Why is website authority harder than publishing content?
Authority depends on outside signals such as relevant links, mentions, citations, reputation, and trust. A site owner can publish content quickly, but earning credible references from other people and organizations takes time and consistency.
What should Double Atari measure as the site evolves?
The most useful measurements are non-branded impressions and clicks, near-miss rankings moving toward page one, relevant referring domains, AI search citations or mentions, consultation inquiries, and engagement metrics.
Should websites be optimized for humans or AI systems?
They need both. The page should be useful, specific, and readable for people, while also giving search engines and AI systems clean structure, clear entities, schema, internal links, and direct answers.
Explore related Double Atari resources: SEO consulting, GEO and AEO optimization, website audits, analytics and reporting, and content strategy.