Bali, Indonesia · [email protected] · linkedin.com/in/widiginanjar
A short overview of how I think about getting found by AI search tools: ChatGPT, Perplexity, and Google's AI summaries, put together for anyone weighing whether this is worth a conversation.
More people now ask an AI directly instead of typing a search query: which vendor to trust, which NGO to support, which product actually solves their problem. That's a different game from ranking on Google. An AI doesn't rank your page, it decides whether to mention you at all, and whether it trusts what it finds enough to repeat it. Most of what breaks that trust is the same structural stuff that's always broken SEO. It just has a new, higher-stakes consequence now.
Search used to be a list of links a person scrolled through. Now it's often a single answer, generated once, handed to the person as fact. If an AI can't find your site, can't parse it cleanly, or finds conflicting signals across your pages, it simply leaves you out, with no ranking penalty to notice and no click-through rate to watch drop. You just quietly don't exist in the answer. That's a harder problem to catch than a slipping Google rank, and most organisations have no way to know it's happening.
Two real audits, anonymised. Both were originally diagnosed as conventional SEO problems, but the same root cause also explains why an AI system would struggle to find or trust these sites.
The situation: Real content, a genuine backlink base, but organic search traffic sitting at 0%.
What I found: Every page on the site shared the exact same title and description, so Google couldn't tell them apart, and neither could anything else trying to read the site. An AI model summarising or citing sources leans hard on exactly these signals (titles, descriptions, distinct page identity) to decide what a page is actually about and whether it's worth quoting. When every page looks identical, there's nothing for a model to differentiate or cite.
Why it matters for AI search: This wasn't just a Google ranking problem. It was a page-identity problem. Fixing it helps human search and AI citation at the same time, because both depend on being able to tell your pages apart.
The situation: Credible testimonials and real published research, but zero visitors from any channel.
What I found: An indexing block left over from staging was quietly telling search engines not to look at the site at all, and no analytics installed anywhere to catch it.
Why it matters for AI search: An AI system that can't crawl a site can't cite it, full stop. This is the most basic gate before any question of "is this trustworthy enough to reference" even comes up. An instruction blocking crawlers blocks every downstream channel, present and future, not just Google's index.
I've spent months as a linguistic reviewer, evaluating AI model responses for factual accuracy. That's given me a working, hands-on sense of what makes a model decide a source is trustworthy enough to cite by name versus something it quietly skips, a different skill from traditional keyword-and-backlink SEO, and one most SEO practitioners haven't had reason to build yet.