The SEO Playbook I Ignored Until My Client's Traffic Tanked

I had a conversation with a client last month that forced me to confront something I'd been quietly avoiding. Their website was ranking #1 for their primary keyword in Islamabad, yet their phone wasn't ringing like it used to. I assumed it was seasonal. Then they showed me their analytics: AI Overviews were now appearing in nearly 70% of their local searches, and when people asked "near me" questions, they saw AI answers 80% of the time. The problem wasn't my ranking strategy. The problem was that ranking well and getting seen by actual humans had become two different things.

That conversation made me realize I'd been optimizing for search engines like it was still 2018. My approach was technical but narrow: crawlable HTML, keyword density, backlink profile. All fine. But I wasn't thinking about how LLMs actually read and cite information. I wasn't building for machines that verify facts across multiple sources, cross-reference data integrity, and synthesize answers. That's a different job entirely.

What Actually Changed in Search

The shift from traditional SERP ranking to AI answer generation means the indexing pipeline is fundamentally different now. Google and other search engines aren't just crawling and returning links, they're extracting entities, verifying that your business data is consistent everywhere, and deciding whether to cite you when an LLM generates an answer.

This matters because an AI engine is roughly 30 times more selective than a traditional search results page. Instead of getting visibility among ten blue links, you're competing to be one of maybe two business names that the model surfaces. That's brutal if you're not built for it.

The second-order effect is that being #1 organically no longer guarantees you'll show up in the AI answer. Your 58% drop in click-through rate on that top ranking? That's because the AI answered the question directly and nobody clicked through to your site.

Structured Data Isn't Optional Anymore

For years, I treated schema.org markup as nice-to-have SEO tax, something to check off before moving on to more interesting problems. I was wrong. It's now the difference between being cited and being invisible.

The technical reality is that LocalBusiness schema with complete, consistent data improves your chances of being cited by roughly 28%. That's not a rounding error. That's the difference between a client thriving and questioning why they're paying you.

What trips most developers up is the consistency part. Your schema data needs to match your Google Business Profile, which needs to match your Yelp listing, which needs to match Apple Business Connect. Mismatched NAP (Name, Address, Phone) data is an immediate trust signal to an LLM that something's wrong. Don't half-ass this.

The sameAs array in your schema is your cross-reference key. It tells the model: "Here's where else you can verify I exist." That's how an LLM confirms it's talking about the same business across multiple platforms.

FAQPage Schema Targets AI-Generated Answers

Here's something I didn't fully appreciate until testing it: pricing and "how does this work" queries trigger AI-generated answers over 80% of the time. Those are also the exact queries where you can directly feed answer content into the model via FAQPage schema.

This is where the technical decision becomes business-critical. If you're building for a service business, you're essentially writing answers that you want an LLM to cite directly. Not answers optimized for human readers, answers optimized for extraction and citation.

My Take: This Reframes What SEO Actually Is

I've been thinking about this wrong. SEO stopped being "optimize for crawlers" years ago, but I treated it like it was until very recently. Now I'm seeing it clearly: SEO is becoming data architecture and information design for multiple types of consumers, humans, crawlers, and LLMs all simultaneously.

What concerns me is the consolidation of trust signals. Review ratings now matter more to AI recommendation engines than they do to traditional rankings. That means your review pipeline, your aggregateRating schema, and your actual review count are now core infrastructure. A business with 200 4.6-star reviews will get cited over one with 15 5-star reviews, all else equal.

The other thing I'd implement immediately: audit your robots.txt file. A surprising number of sites are accidentally blocking GPTBot, PerplexityBot, and other citation-generating crawlers through legacy restrictive rules. You probably want these bots seeing your structured data.

The Checklist for Monday Morning

If you maintain local business websites, here's what I'm implementing this week:

1. Validate your LocalBusiness schema includes aggregateRating, sameAs array, and complete geo data
2. Add FAQPage schema targeting your highest-value pricing and service questions
3. Audit robots.txt for accidental AI bot blocking
4. Set up automated alerts when your aggregateRating differs between your site and GBP

This isn't optional anymore. It's the difference between visibility and irrelevance in search results that are increasingly AI-mediated.

Source: This post was inspired by "AI Search for Local Businesses: How to Compete Beyond Google Rankings" by Dev.to. Read the original article