You know that feeling when you think you finally get how SEO works… and then the internet just… changes the rules overnight? Yeah. That’s kinda what happened to me with this whole AI search engine optimization thing. I remember staring at my screen, watching my perfectly good blog post get buried, and suddenly there’s this weird little AI box on Google giving people answers without them even clicking my site. Like—excuse me?
So, apparently, there’s this new flavor of SEO now. People call it AI SEO, some say AIO, others go on about GEO (generative engine optimization) like it’s the second coming of sliced bread. Honestly? It’s all about making your stuff so clean, so snackable, that AI—whether it’s Google’s fancy Overviews or whatever Perplexity is doing—can grab it, spit it out, and still point people back to you. Sounds easy. It’s not.
And no, it’s not “just SEO but fancier.” GEO vs SEO? SEO’s your classic road trip map. GEO’s like… teaching the GPS to recommend your house as the coffee stop. Big difference. Took me a year (and a lot of ugly traffic drops) to realize if I didn’t adapt, I’d basically be invisible. And trust me, invisibility is overrated.
Section 1: What Is AI SEO (aka AIO/GEO/AEO/LLMO)?
I remember when “searching” meant actually typing something into Google and then… scrolling. Clicking blue links. Getting lost in some WordPress site from 2009 with Comic Sans headings. That was the game. SEO back then? Stuff your page with keywords like “best coffee maker” until it felt like you were writing ransom notes for robots. Worked way too well.
Now? You type anything and an AI just… talks back. Google’s got these AI Overviews—they slap a neat little answer at the top before you even think about clicking. Perplexity does the same. ChatGPT will practically write your term paper before you’ve boiled water for tea. Half the time I forget I even asked. And yeah, if you’re wondering, this is where “AI SEO” comes in—figuring out how to make that voice cite you.
People keep tossing around these acronyms like it’s alphabet soup:
- AIO (AI Optimization) – the broad idea of shaping your content so AI likes it.
- GEO (Generative Engine Optimization) – this one sounds fancy, but it’s just “AI SEO” with a cooler hat. It’s about those AI-powered search tools specifically.
- LLMO (Large Language Model Optimization) – same deal, but focused on getting models like GPT or Claude to pull you in.
- AEO (Answer Engine Optimization) – that’s the older cousin, more about being the “direct answer” in search results before AI showed up to hog the stage.
They all overlap, they all get mixed up, and if you ask five marketers, you’ll get six definitions.
The shift, though—it’s not subtle. The “ranking” game isn’t about just page one anymore. It’s about getting into the AI’s “head” (if you can call it that) so your stuff pops up in that tidy little box where it summarizes everything. And the way you do that isn’t by repeating your keyword like some desperate parrot. It’s by giving the AI little “citation-friendly content blocks.”
I learned that the hard way. Spent weeks writing this 3,000-word masterpiece on home office setups. Thought I nailed it—perfect headings, long-tail keywords, alt tags, the whole checklist. Then I asked Google’s AI Overview about “best ergonomic home office setup” and it quoted… a random Reddit post. Like, word-for-word, some guy named “ChairDaddy47” with two typos and a photo of his cat.
That’s when it clicked. The AI doesn’t care about your 2,000-word essay if it can’t lift a clean, factual, self-contained nugget. Two sentences that answer the question dead-on, maybe a stat, maybe a source, easy to quote. The rest of your genius? The AI will skim it like a bored teenager skimming a syllabus.
So yeah—if you’re thinking about “AI search engine optimization” (see, got it in there), the trick isn’t stuffing more words or obsessing over density. It’s breaking your brain into making little truth bombs the machine can grab. Numbered lists. Quick definitions. Tiny “how-to” steps. Tables. Anything that looks like a complete answer, because to the AI, that’s what trust looks like.
And maybe you’re wondering—does this kill the joy of writing? Sometimes. I mean, it’s weird thinking about robots as your audience. But I’ve stopped writing for them. I write for humans… just in a way that also feeds the robots exactly what they want. Kinda like leaving snacks out for raccoons—do it right, and they keep coming back.
Anyway, that’s AI SEO in my messy little nutshell: blue links are still here, but the spotlight’s moved. You either learn to talk in a way the AI repeats, or you watch “ChairDaddy47” get all the glory. And trust me—you don’t want to lose to that guy twice.
Section 2: How AI Search Engines Work
Alright, so… “how AI search engines work.”
First time someone asked me that, I panicked. Not because I didn’t know — I mean, I kinda knew — but because the way people explain it online? It’s all diagrams and words like “vector embeddings” and “multi-modal contextualization” that make you feel like you’re in a sci-fi movie instead of just… looking for how to rank in AI Overviews.
Here’s the messy, no-buzzword version.
Think of it like this: the AI is basically a nosy librarian with ADHD. It crawls around grabbing everything it can — every page, every blog, every PDF some guy uploaded in 2009 — and dumps it in this giant messy room. That’s the “crawl” part. I once thought my blog posts were invisible because no one “saw” them, but nah… Google’s crawlers had them. They just didn’t care. Painful.
Then it “indexes” the stuff. Imagine the librarian finally sorting your book into the right shelf… except now it’s also judging it. “Hmm… this is about AI SEO, but you barely mentioned Perplexity or entities… so maybe you’re not the expert here.” And yeah, if you don’t have the right structure, it’s basically like handing in an essay without your name on it.
After that? This is where it gets freaky. The AI starts figuring out what your content means. Not just the words — like, if you write “Apple,” it tries to figure out if you’re talking about fruit or Tim Cook’s paycheck. That’s the entity/semantic understanding bit. I messed this up for months — wrote whole posts about “schema” and forgot to mention “structured data” in plain English. The AI thought I was talking about abstract philosophy or something.
Then comes the “answer synthesis” stage. This is when AI Overviews, Perplexity, ChatGPT 5… whatever… take all the sources they’ve found and mash them into a neat, confident-sounding paragraph. Sometimes it feels like they’re that classmate who borrows everyone’s notes, rephrases them, and still gets credit. And you know what? They don’t just grab anything — they pick sources that are clean, fact-rich, and weirdly easy to quote. So if you’ve got one long wall of text with no headings or lists? Forget it. You’re background noise.
And then, finally — if you’re lucky — citations. That’s when your little blue link shows up at the bottom of an AI Overview or inside Perplexity’s answer card. It’s like getting invited to the cool table in high school after eating lunch alone for years. I’ve had it happen, and the traffic spike is ridiculous. But the first time I saw it, I thought it was a fluke. Turned out, the only reason they picked me was because I answered the exact question people were typing in — in one clear sentence — and I used real data, not filler.
Here’s the wild part: the AI isn’t just pulling “top-ranking” stuff. Sometimes it pulls obscure pages… if they answer better. So yeah, you can technically outrank giant sites in AI results without beating them in the classic Google list. That’s why people obsess over how to rank in AI Overviews — it’s a shortcut, but not the lazy kind. You still have to give it a reason to pick you over Wikipedia or some big-name blog.
And Perplexity? Ugh. That one’s picky. It loves citing multiple sources in one answer. If your content doesn’t feel “standalone” and “quote-friendly,” it skips you. I learned this after re-reading one of my posts and realizing there wasn’t a single stat or fact someone could lift without rewriting half a paragraph.
Anyway, all of this to say… AI search engines aren’t magical. They’re just brutal in a different way. They crawl, they index, they figure out “what you actually mean,” they build their little perfect-sounding answers, and then they choose who gets the tiny slice of glory in the citations. And yeah, sometimes they get it wrong, sometimes they skip great content, but if you want in? Keep it clear, keep it structured, keep it worth quoting.
That’s it. No rocket science. Just a nosy librarian who only likes the neat, useful books — and doesn’t care how many hours you spent writing yours if it’s buried in the middle of a rant.
Section 3: Keyword Research in the AI Era
Not the shiny, perfect, “I’ve-got-a-framework” kind. I’m talking about the kind you do at 1 a.m., slouched over your laptop with a half-warm coffee and a browser full of random tabs because you swear you’re just “checking one thing” and suddenly it’s three hours later.
Back in the day — which makes me sound like an old man even though it was, what, five years ago? — keyword research meant hunting exact phrases like “best blender under \$50” and praying you’d rank if you sprinkled it into your post enough times without sounding like a broken robot. And… it worked. For a while.
But AI search? Whole different beast. These things don’t care if you’ve got the exact string “best blender under \$50” twelve times. They want entities. That’s not just a fancy SEO buzzword — it’s basically the “thing” your content is about. The person. The place. The concept. The blender itself. And all the bits and bobs around it. Attributes, relationships, how it connects to other “things” in the web’s massive brain.
Think of it like gossip. The AI’s not just asking “who’s John?” It wants to know who John’s dating, where he lives, what his weird hobby is, and whether he’s the guy who wore socks with sandals to the wedding. The more context, the better.
So instead of obsessing over one rigid keyword, I start with a core entity — say, “Perplexity AI” — and then branch into everything tied to it. Sub-entities: “Pro subscription,” “AI citation sources,” “mobile app features.” Attributes: price, release date, citation accuracy. Then? Questions. Not the fake “SEO-verified” ones, but real ones people blurt out online like is Perplexity better than ChatGPT for research? or why does it keep citing Medium blogs?
And that’s where the “question mining” rabbit hole happens. Google’s People Also Ask boxes are gold. Forums are even better because people are unfiltered there. Reddit threads, Quora answers, random YouTube comments — you find the actual stuff folks type when they’re mildly annoyed or genuinely curious.
If you’re thinking “cool but I’ll get lost,” yeah, you will. But that’s where clusters come in. You grab your core entity and start mapping:
| Entity | Attributes | Common Questions | Content Block Idea |
|---|---|---|---|
| Perplexity AI | Pro subscription, mobile app, citation style | How to get cited? Is it better than ChatGPT? Does it use GPT-4? | Side-by-side feature comparison table |
| AI Overviews | Sources, update frequency, snippet length | How to appear in AI Overviews? Do images help? | Step-by-step “AI Overview Optimization” checklist |
| Entity SEO | Entity attributes, schema types, linking | What’s entity SEO? Is it still relevant? | Beginner’s guide with diagrams |
| Topic Clusters | Pillar pages, subtopics, linking structure | How to build topic clusters? | Visual cluster map + linking tips |
Now here’s where I screwed up for years — I used to think “more keywords = more reach.” Nope. AI search engines sniff that out. They want “information gain,” which basically means: are you saying something new or are you just copy-pasting the internet with prettier fonts?
So if the top 10 results all say “add FAQ schema,” maybe you show how adding it bumped your impressions by 37% in two weeks. That’s the kind of detail AI surfaces love because it’s unique.
And yeah, I’ve had nights where I stare at a keyword list and think “there’s nothing here” — but then I throw weird prompts into ChatGPT like:
- “List 20 common misconceptions about [entity].”
- “Pretend you’re a customer angry about [entity] and list 10 complaints.”
- “Write fake Reddit thread titles about [entity].”
Half of it’s trash. The other half? It’s the stuff nobody else is covering.
You don’t have to overcomplicate it. Core entity → attributes → real questions → your own spin. Wrap it in content blocks AI can easily lift (short definition, table, numbered list, stat box). Make sure it actually says something useful, not just what’s “SEO correct.”
And honestly… sometimes I write for the AI and forget humans will read it too. That’s a mistake. Because if it’s boring, they’ll bounce, and guess what? AI notices that too.
So yeah. Build clusters, mine questions, hunt for that one angle everyone’s missed. And if it feels messy? That’s fine. The best maps are scribbled on napkins first.
Section 4: Content Architecture for AI Citations
Alright, so—content architecture for AI citations.
This one took me way too long to figure out. And not because it’s complicated in some “rocket science” kind of way… but because I used to overthink every single line I wrote. I’d sit there rearranging sentences like some deranged librarian, thinking this would make Google love me. Spoiler: it didn’t. AI didn’t care. Humans didn’t care. My cat didn’t care.
What actually matters? Making it stupidly easy for an AI to grab your answer and throw it into one of those shiny little boxes at the top. You know, the ones where it spits out two or three sentences and then maybe—maybe—drops your link. If you want in, you gotta feed it exactly what it can digest without choking.
Here’s what I learned the hard way:
1. Start with the answer.
Not a poetic intro. Not a cliffhanger. Just… the damn answer.
Two to four sentences. Like you’re explaining to a stranger in an elevator before the doors open. “What is [thing]? It’s X, it works like Y, and here’s the most important thing you should know.” Done. Then you can ramble below if you want. AI loves that clean upfront definition because it doesn’t have to dig through your nostalgia about learning SEO in a cyber café in 2009.
I remember the first time I rewrote an old post this way—just slapped a neat definition at the top, then broke the rest into bite-size chunks. Two weeks later, I saw my words inside Google’s AI Overviews. My first reaction? “That’s… cool? But also, hey, could you maybe show my site name bigger?”
2. Give it blocks, not walls.
AI hates walls of text. Humans do too, but AI is way more unforgiving. Numbered steps? Tables? Those little “Pros” vs “Cons” side-by-sides? Gold.
Like—imagine you’re writing about, say, optimizing your schema markup. Instead of burying it in paragraphs, do:
| Step | Action | Why it Matters |
|---|---|---|
| 1 | Add FAQPage schema | Helps AI understand your Q\&A format |
| 2 | Validate with Rich Results Test | Prevents broken markup from being ignored |
See how easy it is for a bot to lift that? You’re basically gift-wrapping it.
And stat callouts—oh man. Put numbers in bold. Add the source in parentheses. “Sites with structured schema are 27% more likely to appear in AI Overviews (Search Engine Land, 2024).” That’s the stuff that gets scraped because it’s self-contained.
3. Schema is like subtitles for your site.
Don’t skip it. I used to treat schema like flossing. You know you should… but eh, maybe tomorrow. Big mistake. AI search engines are basically schema junkies. Feed them Article, FAQPage, HowTo, Product, Organization, Person… whatever fits. And if you’ve got a nice tight summary? Try Speakable. It’s like telling the bot, “Hey, this part is quote-ready.”
Oh, and if you’ve got multiple authors or an actual company behind the content? Use Person and Organization. Give them bios, social links, proof you’re real. Which leads me to—
4. E-E-A-T isn’t just a Google buzzword.
It’s basically “Hey, prove you’re not some faceless content farm.” I slap my name on stuff now. Add my own photo. Put “Last updated: [date]” like I’m showing my homework. If I have original data—screenshots from my analytics, a survey I actually ran—I use it. Because AI engines seem to reward the stuff that feels anchored in a real human life, not just a rewrite of someone else’s rewrite.
And weirdly enough? Readers notice too. I once posted a chart I made in Canva with survey results from exactly 27 people (yes, small sample size, I know), and someone emailed me to say, “Thanks, this was the first time I saw actual numbers from a real person.” That hit me.
5. Don’t make them guess.
By “them,” I mean both the bots and the people. Label things clearly. Headings that tell the whole story—“Steps to Optimize FAQPage Schema” is better than “Tips.” Keep alt text descriptive but short. And if you link out, link to authoritative stuff… even if it’s a competitor. AI notices credibility signals.
Sometimes I imagine AI as a slightly impatient intern. You either hand them a clean, labeled folder… or you dump a shoebox full of random papers on their desk. Which one do you think makes it into the presentation?
I guess what I’m saying is… this isn’t about “tricking” AI. It’s about making your stuff so clean, so digestible, and so obviously human-made that the machine can’t not use it.
When I stopped writing to impress other writers and started writing like I was building a LEGO kit for a lazy robot—step by step, clear labels, bright colors—that’s when I started seeing my content pop up in AI results.
So yeah. Answer first. Break it up. Give it schema. Show your face. Drop some numbers. And… maybe floss your teeth while you’re at it. It’s all connected somehow.
Section 5: On-Page Optimization (2025 Edition)
I’ll be real with you — “on-page optimization” sounds like the sort of thing I used to skip because I thought it was just… boring. Like, “change your H2s, add some alt text, fix your links” — yeah, sure, whatever, next. But then one day I actually looked at why half my pages were invisible to both Google and these shiny new AI search things. And it hit me. It wasn’t my ideas. It wasn’t my writing. It was because my stuff was basically invisible in all the right places.
So now, headlines? I treat them like little neon signs. None of this “cute but vague” stuff I used to write at 2 a.m. “Why Backlinks Are Cool” — nah, that’s dead. Now it’s “Why Backlinks (Link-Building in 2025) Still Decide Who Wins.” See what I did there? Entity in there (“link-building”), year in there, direct promise. And it’s not just H1s — H2s, H3s, even the random subheads I used to just bold — they all carry some sort of thing the search engine (and yeah, the AI overview bot) can latch onto.
And skimmable? I learned the hard way when a friend messaged me like, “I read your post but… I couldn’t find the answer so I asked Perplexity instead.” That was a knife to the gut. So now every chunk has a label. If it’s “Step 3: Compress Your Images” — that’s the subhead. Not “Fix This Before It’s Too Late” (which sounds cool, but useless to a machine).
Then there’s alt text. Oh boy. My early alt texts were tragic — “image123.png” or worse, blank. That’s like showing someone a photo in the dark. Now I make them… descriptive, but not robotic. If I post a screenshot of my analytics? It’s “Google Search Console showing AI Overviews clicks for blog post on entity SEO” — long, yes, but the bot eats it up. Same with filenames. If the image is about topical authority, the file isn’t “chart-final-final2.png.” It’s “topical-authority-internal-links-diagram.png.” It’s work, but the day I saw my image show up in a chat answer? Yeah. That was a little dopamine rush.
Internal links though — man, I used to be reckless. Linking “click here” and “read more” like it was 2009. Then I realized those links are basically breadcrumbs for search engines and AI scrapers. So now if I’m talking about structured data, the link says “structured data schema guide” and it goes to my own schema page. And I don’t go nuts linking everything — just enough so if someone (or a bot) is crawling, they see I have a whole neighborhood on the topic, not a bunch of dead ends.
Charts, diagrams, captions… yeah, captions are underrated. People skim them, bots read them, and if you drop a keyword or an entity there, it sticks. And don’t make the captions boring — if you’ve got a chart, say what’s in it and why it matters, not just “Figure 2.” That’s cold.
And I can’t forget the FAQ box. I fought using it at first — I thought it made my pages look like some cheap affiliate site. But guess what? AI loves them. They’re basically pre-packaged answers for the bot to grab. I pull my questions from actual searches — those weirdly specific ones like “does FAQ schema work for AI Overviews” or “how many internal links per 1,000 words is too many.” I write them short, answer in 2–3 sentences, drop schema markup, done. I swear I’ve had answers pulled into AI Overviews verbatim.
Anyway… I’m still not perfect at this. I’ll probably look back in a year and cringe at half of what I’m doing now. But if I can give you one thing: treat every heading, every alt text, every internal link like you’re making a map for both a human and a machine. One’s impatient, the other’s dumb — but they both need you to be clear. And yeah, maybe I overthink filenames now, but if “topical-authority-internal-links-diagram.png” gets me even one more citation in an AI search result, I’m cool with it.
Section 6: Technical Foundations That Still Matter
Alright, so… “technical foundations.”
Yeah, I used to roll my eyes at that too. Sounds like something a corporate SEO manager says right before showing you a 48-slide deck you pretend to read while scrolling Instagram under the table. But then, one day, I nuked an entire site’s rankings because I ignored this stuff. And not in a fun “oops haha” way — in a “why is Google pretending my site doesn’t exist anymore?” way.
Crawlability first. You ever lock yourself out of your own house? I did, twice, because I left the key inside. That’s what a robots.txt file can do if you mess it up — you’re basically telling Google, “No, don’t come in, I’m busy,” and then wondering why nobody shows up. So yeah, make sure Google can actually get to your pages. Not just your homepage, but the weird deep ones too. The ones you forgot you even had.
Then there’s “clean IA” — information architecture, but honestly, it’s just “don’t make your site a maze.” People (and bots) shouldn’t have to click six mystery links to find your “Contact Us” page. I once had a menu so confusing my own mom couldn’t find my about page. She called me to ask if I actually ran the site or if it was a scam.
Canonicalization… god, I learned this the hard way. Had the same page living on three URLs because of some sloppy CMS setting. Guess what? Google didn’t know which one was “the one” and basically gave up. So yeah, pick one version, mark it, and stop confusing the poor bot.
Core Web Vitals? Look, I’m not a speed freak. But when your site takes five seconds to load on mobile, people will leave before your hero image even fades in. I had a page with a beautiful high-res photo of a sunset. Only problem — it was 4MB. On mobile, that’s a death sentence. Shrink it, lazy-load it, whatever. Nobody cares how sharp your clouds look if they never see them.
Speaking of mobile… if your site still looks like it’s from 2010 on a phone, I don’t care how great your SEO is, you’re toast. Google will notice, and people will bounce. I’ve been there, watching analytics show “0:03 average time on page” like a slow-motion car crash.
HTTPS — yeah, secure your damn site. It’s not 2013 anymore. I once left a site on HTTP because “I’ll deal with it later” and then my browser started slapping a huge “Not Secure” label on it. Kinda kills the vibe.
Structured data validation? This one’s sneaky. I thought I had schema set up right. Then I ran it through Google’s tool and… red errors everywhere. JSON-LD is picky. Missing commas, wrong property names — one tiny typo and it’s like the bot goes, “lol nope.” Fix it. Test it. And XML sitemaps too — they’re like a to-do list for Google. But keep them clean. No 404s. No random junk.
Thin or duplicate pages — just delete them. Seriously. I once kept 40 “placeholder” pages live thinking they’d help. Nope, they just made my site look lazy. And pagination… make it make sense. If Google gets lost in your /page/4/ → /page/5/ loop, you’re wasting crawl budget.
UGC (user-generated content) — comments, forums, whatever — can be gold, but it’s also a spam magnet. I had a blog comment section full of “Nice post! Buy shoes here → shady.link” and Google hated me for it. Moderate or block it.
So yeah… technical SEO isn’t glamorous. It’s basically cleaning your house so guests don’t trip on a Lego and sue you. But if you skip it, AI search or not, your content won’t even get in the door. And trust me — fixing it after the fact is like trying to untangle headphones that’ve been in your pocket since last summer.
Section 7: GEO / LLM Optimization Tactics (Google, Perplexity, ChatGPT)
Alright, so—GEO, LLM optimization, all that shiny buzzword soup—people make it sound like some mysterious monk-level skill. It’s not. It’s messy. And yeah, I’ve screwed it up before.
I remember the first time I tried to “rank” in Google’s AI Overviews. I thought, okay, just crank out a fat keyword-stuffed post, like old-school SEO but fancier. Nope. Didn’t even make it into the AI blurb. My site was invisible. Turns out, Google’s AI doesn’t care how many times you say “geo optimization.” It wants neat, bite-sized facts it can steal and pretend it thought of itself.
You know that feeling when you’re talking to someone who interrupts you to summarize what you just said? That’s AI Overviews. If your content isn’t organized into tiny, snackable blocks—2–3 sentence definitions, step lists, mini-tables—it just… skips you. Worse, if you repeat yourself or pad the copy? Dedupe hell. The AI sees it once, assumes the rest is redundant, and ignores you.
So what works?
- Lead with the answer. Literally the first 2 sentences: “X is Y. Here’s why it matters.”
- Slip in a stat or fact they can’t find anywhere else. Not “according to a report” generic stuff—like something you learned from your own test, your own data. That’s “original insight” in AI land.
- Show you’ve actually touched the thing you’re writing about—screenshots, personal results, ugly mistakes. Google’s “experience signals” basically mean: prove you didn’t just rewrite Wikipedia while eating cereal.
Then there’s Perplexity. Different beast. More curious, less judge-y. It likes pulling from multiple sources in one answer, so if you’re the only site saying something? Cool. But if all you’ve got is what everyone else has? You’re just another tree in the forest. I started sneaking in fresh examples and updating old posts weekly. Perplexity loves recency. It’ll even cite some random blog post from last Tuesday if it’s clear, well-structured, and has the author’s name right there. Yeah—real name, not “admin.” I lost a citation once because I didn’t have my author box set up. That’s how petty it gets.
And ChatGPT’s search-style answers? Man… they’re picky. It’s all about structure. If you’ve got one massive wall of text, forget it. Break it down like you’re leaving sticky notes for a forgetful roommate:
- Definition
- 3–5 bullet points
- Quick checklist
- One-sentence summary at the end
Also, don’t be boring. The AI will choose something else. I once ran two similar posts—one dry, one with a dumb joke about pizza—and guess which one got cited? Yeah. Pizza wins.
Alright, I promised a side-by-side. Here’s the ugly truth of “Old SEO” vs “AI-ready”:
Old SEO Page:
Title: “10 Ways to Improve Your Website SEO”
Intro: 300 words of warm-up fluff
Body: Long paragraphs, mixed tips, random tangents, no clear structure
No author bio, no schema, no updated date, no original data
Calls to action: buried at the end, after you’ve already lost people
AI-Ready Page:
Title: “10 Proven Ways to Improve SEO (Updated Feb 2025)”
Intro: 2 sentences explaining what + why
Each tip = mini-block (definition → steps → example → source)
Schema: FAQPage + HowTo + Person (author) + last updated date
Original chart comparing before/after results
Bullets, tables, stats pulled from your own data
Internal links to deep-dive pages
CTA right after the intro, another in the middle, one at the end
See the difference? The old page was a story. The new page is a vending machine—AI walks in, grabs what it wants, and leaves your brand name in the “Sources” list like a little calling card. That’s the whole game now: be quotable and citable.
Do I love it? No. I miss writing for humans first. But also—humans aren’t finding your site if you’re invisible in AI results. So you do both. You make it snackable for the bots and spicy enough for the people.
Anyway, I’m not saying you’ll hit #1 in AI Overviews tomorrow. I still get blanked sometimes, and it stings. But I’ve seen pages go from zero to sitting inside Perplexity’s top three sources in a week… just by tightening the blocks, refreshing the date, and making sure my dumb face + name were right there.
If you want, I can give you my “GEO block” template that I just copy-paste into everything now—it’s stupid simple, but it works.
Section 8: Brand, PR, and Link Signals in an AI World
Okay, so, brand mentions.
God, I used to roll my eyes at that stuff. Like, who cares if someone name-drops you without a link? In my head, if it wasn’t blue and clickable, it didn’t matter. I was chasing backlinks like a raccoon after shiny things — guest posts, directory listings, begging some old blog to “please link my resource.” And then… AI search shows up and flips the table.
Because now? Half the time the AI pulls your stuff without sending traffic, and the only clue you were even “there” is your name showing up in its citations. Or not even citations. Just… your brand floating in that weird AI-generated paragraph like a ghost in someone else’s house. It’s unsettling. But also… kind of powerful.
I had this one week where Perplexity (yeah, I stalk my brand mentions like it owes me money) cited my site three times — but only one had a link. The others? Just the name. And you’d think, “meh, waste.” But here’s the thing — next week, a completely different site writes about the topic, and they link. Where’d they find me? I’m betting they saw my name in that AI answer. Which means… apparently people do Google stuff they see in AI blurbs.
So yeah, “brand mentions ai search” — it’s not just an SEO buzzword now, it’s currency.
And the creepy part? AI engines seem to trust you more if your name shows up in a lot of credible places. Like it’s building some weird, invisible web of, “Oh, this brand keeps coming up in trustworthy sources, so let’s feed it into more answers.” Which means PR suddenly matters more than my old keyword spreadsheets.
And I’m not talking red carpets or press releases with stock photos of handshakes. I mean digital PR that actually has teeth — original research, data no one else has, surveys that get picked up because they’re oddly specific. One of my friends runs a gardening blog (don’t laugh, it’s surprisingly competitive), and she published a small study on how often potted plants survive when watered with cold vs. room-temp water. She got quoted in Better Homes & Gardens. Guess who’s in AI Overviews now for “best way to water plants”? Yep.
I’ve screwed this up before, too. I once “did PR” by blasting out a half-baked “industry trends report” that was basically just me paraphrasing other people’s data. No one cared. No one linked. And I’m 90% sure Google’s AI laughed at it. The stuff that sticks is the stuff you’ve actually done. Ran the experiment. Collected the numbers. Got dirty.
Surveys, weird polls, niche experiments… that’s the stuff journalists pick up. And once you’re in their article? Boom — your brand exists in a credible context. AI eats that up. It’s like… not just links, but little trust crumbs scattered all over the web.
And yeah, links still help. Don’t throw them out. But I’ve stopped obsessing over “link juice” like it’s 2015. Now I care more about whether my name is sitting next to other names the AI already loves. If a university site says my name, if a local news piece quotes me, if my data shows up in a chart on a .gov page — that’s the good stuff. That’s how you earn citations in ai results without screaming for them.
Honestly, it’s slower. And less measurable in the short term. And it’ll make you wonder if you’re wasting time. But when your brand starts popping up in AI search — even without the link — you’ll know. You’ll feel it. And you’ll realize this isn’t about chasing algorithms anymore… it’s about becoming part of the story the web tells when it’s asked a question.
Anyway, that’s what I’ve been thinking about. And yeah, I still chase shiny backlinks sometimes. But now I chase mentions like they matter just as much. Because they do.
Section 9: Measurement & Tooling
Whatever you wanna call it.
This is the part nobody gets excited about, but it’s the thing that quietly decides whether all the hours you spent tweaking titles and shoving schema into your posts were worth anything or just a really elaborate hobby.
I remember when I first heard “track your AI citations.” I thought, okay… cool? But then I realized, uh, nobody told me how to do that. It’s not like Google Search Console suddenly grew a little “AI Overviews” tab. I mean, maybe one day… but right now it’s like trying to hear your name in a crowd when everyone’s wearing headphones.
Here’s what I started doing. And yes, it’s messy.
I keep a spreadsheet (don’t laugh, it works) where I jot down whenever I spot my stuff in AI Overviews, Perplexity, even random ChatGPT answers my friends send me screenshots of. Yeah, it’s low-tech, but at least it’s a start. I track:
- AI citations (the exact page they linked to, the snippet they pulled)
- Answer-box appearances (because, let’s be honest, if you’re in those, you’re halfway to AI results anyway)
- And this weird one — “assisted traffic.” Basically, I look at my analytics for sudden traffic spikes with no obvious referrer. It’s usually AI search quietly handing people my stuff without a click trail. Creepy but… thanks, I guess?
Now, tool-wise… it’s like the wild west. Some AI SEO analytics tools claim they can tell you every time you’re mentioned in AI search results. Some are solid, some feel like they were built in someone’s garage last weekend. I’ve tried:
- AI citation trackers (they scrape or simulate AI searches to see if you’re cited — hit or miss)
- Entity extractors (geeky little tools that pull out “entities” from your content so you can see if Google even knows what you’re about)
- Schema validators (because broken schema is like wearing your shirt inside out — technically fine, but people will look at you weird)
- And yeah, boring stuff like log-file analyzers — not fun, but they show if bots are actually hitting the pages you care about.
If you’re wondering, “Do I need all of that?”… nah. Start with one or two. Or none. Just… don’t do what I did and assume “I’ll just check later.” Later never comes. And when you finally look, you’ll wish you had the receipts.
The dashboard thing… okay, this is where I made a total mess at first. I wanted this clean, minimalist, beautiful thing that would make me look like I had my life together. Instead, I ended up with a Franken-dashboard in Google Data Studio that had 17 different charts and gave me a headache every time I opened it.
So now? It’s stupid simple:
- KPIs: AI citations count, % change month over month, answer-box hits, estimated assisted traffic.
- Cadence: once a month check-in, no daily obsession (that way lies madness).
- One sheet. One chart for each thing. If it takes longer than 10 minutes to update, I did it wrong.
And yeah… I’m still figuring it out. Some months I think I nailed it, others I’m like, “Wait, did AI just ghost me?” But at least now I can tell when something’s working instead of just guessing.
Anyway. Measure it. Even if it’s ugly. Even if you feel like you’re wasting time. Because the day your AI SEO starts actually working and you have proof — real proof — you’re gonna be glad you weren’t just trusting your gut.
Section 10: Governance, Ethics & Risk
I’ll be real with you.
Half the “AI SEO ethics” talk online feels like someone stapled a PR statement to a buzzword list. I’ve read those posts, nodded along, and then… went right back to doing stuff I wasn’t proud of. Not evil stuff. Just lazy. Like, letting an AI spit out a fact without checking it because ugh… I was tired, the deadline was stupid, and who’s even going to notice, right?
Except they do notice. Not right away, maybe, but one day some guy emails you at 2 AM saying, “Hey, this stat you quoted? Yeah, it’s from 2012 and completely wrong now.” And you sit there, phone glowing in the dark, feeling like you just got caught copying homework. That’s when “accuracy” stops being a bullet point on a slide and starts being… this knot in your stomach you don’t forget.
Source attribution? That’s just a fancy way of saying don’t steal. If you grabbed a number from some researcher in Finland who spent two years freezing their butt off to collect it, give them the credit. Link their name. Drop their site. And maybe… idk… read the actual source before you mash it into your post. I once cited a blog that turned out to be quoting another blog that was quoting Reddit. Like, a game of telephone, but the ending was “everyone’s wrong.”
Bias? Yeah, we all have it. Even AI. Especially AI. I’ve asked AI to write about “best SEO tools” and it kept naming the same five big ones — like there aren’t fifty indie ones that are cheaper and better. Sometimes the bias is baked in, sometimes it’s just… inertia. If you don’t have a human review what the AI spits out, you’ll never see the blind spots. And those blind spots are where your audience decides, nah, this isn’t for me.
Compliance sounds boring, and yeah, it is. But it’s also the thing that keeps you from waking up to a “cease and desist” email. GDPR, copyright, FTC disclosures — it’s all that junk you think doesn’t apply to your little site until it does.
And human review? I used to think it was just a box to tick. But the more I work with AI, the more I realize it’s like… spot-checking your parachute before you jump. If you can’t document what you checked — where the facts came from, why you kept or cut something — you’re basically freefalling and hoping it all works out.
So yeah. AI can help you move fast, but ethics is what keeps you from faceplanting in public. And once you screw that up, people remember. Longer than your traffic spike, longer than your “viral moment,” longer than you’ll probably like.
Section 11: Case Studies & Swipe Files
This “case studies & swipe file” thing — I’ve got a weird relationship with it. Feels kinda fake when people only show the perfect wins, right? Like, look at my chart go up — and they never mention the six months where everything was flat and you were refreshing Search Console like a maniac. So I’m just gonna tell you how it actually went down for me and a couple of people I know (names changed, because I like having friends).
Case 1: The “Oh Crap, I’m Invisible” Blog
This guy — let’s call him Sam — had a site that was… fine. Pretty, clean, words on a page. The problem? AI Overviews treated him like he didn’t exist. I saw his page in the top 5 on Google classic search but in AI’s answer box? Nope. Not a peep.
We stripped it down. Like, tore the thing apart. Turned chunky paragraphs into these short, punchy “answer blocks.” Each block basically answered one hyper-specific question in 2–4 sentences. Added a little table here, a bulleted step-by-step there.
Before:
<h2>How to Make Cold Brew Coffee</h2>
<p>Cold brew coffee is a method where you steep coarse ground coffee in water for a long time...</p>After (yes, this is swipeable):
<h2>Cold Brew Coffee: Quick Answer</h2>
<p>Cold brew coffee is made by steeping coarse ground coffee in cold water for 12–24 hours.
Use 1 cup coffee beans for 4 cups water.</p>
<h3>Steps</h3>
<ol>
<li>Grind coffee coarse.</li>
<li>Mix with cold water.</li>
<li>Cover & steep 12–24 hours.</li>
<li>Strain, chill, serve.</li>
</ol>We slapped on FAQPage schema for good measure, hit publish, and two weeks later? He’s cited in AI Overviews and Perplexity. Traffic bump: +38% in a month. Not life-changing, but enough to make him stop cursing at his analytics.
Case 2: My “Embarrassing Table” Moment
I once spent days making this gorgeous infographic for a guide on “best budget cameras.” It was beautiful. Looked like something from a tech magazine. Problem was… AI search engines don’t care how pretty it is if it’s stuck in an image file. They can’t “see” it. Rookie mistake.
So I went back — embarrassed, annoyed — and made this ugly HTML table instead. It looked… fine. But machines? They loved it.
Here’s the pattern (feel free to steal):
<table>
<tr><th>Camera</th><th>Price</th><th>Resolution</th><th>Best For</th></tr>
<tr><td>Brand X100</td><td>$399</td><td>20MP</td><td>Travel</td></tr>
<tr><td>Brand Y20</td><td>$299</td><td>16MP</td><td>Beginners</td></tr>
<tr><td>Brand ZPro</td><td>$499</td><td>24MP</td><td>Vlogging</td></tr>
</table>After that, I got cited in Perplexity for “best budget cameras under \$500.” It still makes me laugh. All that design work… beat by a boring HTML table.
Case 3: FAQ Bombing the Niche
Friend of mine — she runs a gardening blog — was barely holding on in rankings. She started writing these tiny FAQ sections at the bottom of posts. Simple stuff: “How often should I water basil?” “Can basil grow indoors?”
Each one got its own <h3> with a direct answer in one or two sentences. Then we wrapped the whole thing in FAQ schema.
Pattern:
<h3>How often should I water basil?</h3>
<p>Water basil every 2–3 days, keeping the soil slightly moist but not soggy.</p>Fast forward a month… she’s not just ranking in Google snippets, she’s popping up in AI-generated answers. Not the top traffic spike of all time, but the quality of visits? Way better. People were staying longer, actually clicking around.
Quick Swipe File — Steal These Patterns
Definition Block
<h2>[Term]: Quick Definition</h2>
<p>[1–2 sentence clear definition with key numbers, facts, or timeframe.]</p>Steps Block
<h3>Steps to [Action]</h3>
<ol>
<li>Step one — keep it short.</li>
<li>Step two — add numbers if possible.</li>
<li>Step three — finish with result.</li>
</ol>Pros/Cons
<h3>Pros</h3>
<ul><li>Pro 1</li><li>Pro 2</li></ul>
<h3>Cons</h3>
<ul><li>Con 1</li><li>Con 2</li></ul>Comparison Table — see my “embarrassing table” above.
FAQ Block — small, direct, 1 Q + 1 A per <h3> and <p> pair.
Look, I’m not gonna pretend this is magic. Sometimes you do all this and nothing happens. Sometimes AI Overviews ignore you. But… these patterns? They’re low-effort enough that if they work, you win. If they don’t, you haven’t lost weeks of your life.
If you want, I can also give you the exact schema JSON-LD I’ve been using in these wins — the same one that got my cold brew and budget camera posts into AI Overviews in under 3 weeks.
Section 12: FAQs
Not the shiny marketing kind, just the questions people actually DM me or blurt out at meetups when I’m halfway through a sandwich.
Q: Is SEO dead because of AI?
Short answer? Nah. Long answer… okay, I thought it was for a hot minute. Like, when AI Overviews started rolling out and Perplexity was spitting out perfect, polished answers without anyone clicking through, I sat there thinking, cool, so I just wasted the last 8 years of my life learning how to rank for blue links. But here’s the thing—someone still has to feed the AI. It’s like when your younger cousin suddenly “knows everything” because they read it on Wikipedia. Where’d Wikipedia get it? People. Writers. Sites. That’s you, if you’re smart about it. So no, it’s not dead. It’s just… moved house.
Q: How long to get cited in AI Overviews?
Honestly? I wish I could give you a sexy number like “37 days.” I’ve had one post get picked up in two weeks because it had a weird stat nobody else mentioned. I’ve also had stuff I knew was gold just sit there, ignored, for months. AI models are like those professors who take forever to grade your paper but still drink your milk from the staff fridge—there’s no guaranteed timeline. Best you can do? Make your stuff too good (or too oddly specific) for them to ignore. And yeah, sometimes you’re just yelling into the void until one day, bam, your screenshot’s in an AI answer.
Q: Do I still need links?
Yeah. Sorry. I know some AI gurus swear “links are dead,” but every time I’ve had something blow up in search—whether human-click or AI citation—there’s been a trail of decent backlinks behind it. It’s like, AI still uses the internet’s old trust signals, they just don’t brag about it. And I’m not talking 500 spammy directory links from 2009. I mean someone with an actual audience mentioning you because you had something worth pointing at. Bonus points if they weren’t your mom.
Q: GEO vs traditional SEO — do both?
If you’ve got the time, yeah. I mean, GEO (generative engine optimization) feels like prepping your notes so the smartest kid in class copies from you during an exam. Traditional SEO is still about owning your spot on the page where people can choose to click. Why pick one? You can double your chances. But—if I’m being real—some days I only have energy for one, and I still lean old-school because clicks = traffic I can actually measure. GEO traffic’s still a bit ghosty, you know? It’s there, you just can’t always prove it to the client unless you’ve got screenshots.
Anyway, point is, AI’s not killing SEO, it’s just… mutating it. Like that old dog that suddenly learns a new trick but still needs the same old food. Keep feeding it. Keep tweaking the recipe. And maybe, once in a while, yell at it from the porch just to remind it who’s boss.
Conclusion + Action Plan
I’ll be honest, this whole “AI SEO strategy template” thing sounds fancy on paper, but in my head it’s just… five things I keep scribbling on sticky notes and losing under my keyboard. Research entities — that’s step one, because if you don’t know the “things” people and machines care about, you’re just throwing words into the void. I’ve skipped this before, thought I could wing it… ended up with a 2,000-word rant about nothing. Nobody read it.
Then you structure blocks. Not like Lego, more like… bite-sized bits that an AI can grab without choking. I’ve made the mistake of hiding the good stuff in paragraph 12. Big no. Put it in their face.
Add schema — yeah, the part that feels like coding but isn’t. I used to ignore it because I thought it was “extra,” but turns out the robots like neat labels.
Strengthen brand mentions. Ugh. I hated this step at first because self-promotion feels gross. But if people (and AI) don’t see your name floating around, you’re invisible.
And measure AI citations… my least favorite, but also the only way to know if any of this works. Otherwise, you’re just poking at a dead engine hoping it roars.
Do those five, in order, over and over. Not perfect. Just… keep looping. That’s my messy “AI SEO roadmap 2025,” for what it’s worth.