SEO and AEO, explained without the fog
Written by the team that sells the work, which cuts both ways: every guide names real methods and real prices, and none of them pretends the answer is "it depends" when it is not.
What is AEO?
Answer Engine Optimization, defined. SEO earns you a ranking; AEO earns you a citation inside the AI answer a buyer actually reads.
llms.txt: an hour of work, honestly assessed
What it is, the honest state of adoption, and how to write one. Includes our own as a working example.
How to get cited by ChatGPT
Five mechanical moves in order of leverage, from letting the crawlers in to monitoring the drift.
AI Overviews: how Google picks its sources
Rank first, structure second, monitor always. The overview mechanics without the mysticism.
Entity SEO: become something engines understand
Why engines reason about entities, not keywords, and how to make them certain who you are.
The AI crawlers, and how to let them in
GPTBot, ClaudeBot, PerplexityBot and Google-Extended, and the robots.txt that lets the right ones through.
Schema markup that earns AI citations
The six schema types that matter, why they move citations, and the one rule that keeps you safe.
GEO, AEO, SEO: the acronyms, sorted
What each means, how much is a rebrand, and what actually changes about the work.
How to get cited by Perplexity
Two crawlers with different rules, conflicting official guidance, and the one AI engine whose citations show up in your analytics.
How to measure AI visibility
Four instruments, what each one proves, and why Search Console now reports AI impressions but will never give you clicks.
Optimizing for Google AI Mode
One question becomes a dozen searches you never see, and Google says there is no special markup for any of them.
Why is my traffic down?
The five real causes of an organic drop, in the order you should check them, before you fix the wrong thing.
How long does SEO take?
Three to six months for real movement, and the three factors that decide where you land in that range.
Why is my competitor outranking me?
The technical, content and authority gaps that let a weaker page win, and how to find out which one is yours.
Website migration without losing rankings
Three official Google pages give three different answers on how long to keep your redirects, and the one asset you cannot recreate after launch.
Local SEO and AI search
The map pack runs on geography and the AI answer runs on citations. Why a perfect Business Profile wins one system and does almost nothing for the other.
Do backlinks work on AI search?
One study puts link authority against AI visibility at 0.65, another at 0.27, and nofollow links track AI mentions as closely as follow links do.
How much does SEO cost?
Real pricing ranges by business size, the three ways agencies charge, and the guarantee-based red flags that mean you are being overcharged.
How to get cited by Microsoft Copilot
It answers from Bing's own index. Bing now reports the citations, the intents and the citation share, and Ahrefs shows exactly who is winning them.
Keyword research for AI search
Semrush found 65 to 85% of AI prompts match no traditional keyword, and Ahrefs found only 12% of AI citations rank in Google's top 10 for the same prompt.
Is AI search traffic worth it?
A quarter of a percent of the average site's traffic, converting at up to 23x. Both numbers come from SEO tool vendors, and here is the arithmetic for yours.
Do AI crawlers render JavaScript?
No, and a 41-day experiment found GPTBot and ClaudeBot reached exactly zero pages linked by JavaScript. Your Google ranking proves nothing about this.
Does on-page SEO still matter?
A 2025 study found Google rewrites 76% of title tags and keeps only 35% of the original wording. What still moves the outcome: internal links, schema and content that matches what already ranks.
How much content should you publish?
An agency dataset of 14,000 posts found quadrupling output made rankings worse. What predicts an AI citation instead: the update date, not the publish date.
Does rank tracking still matter?
A June 2026 study found 68% of Google searches end without a click, and AI Overviews cut click-through nearly 60%. What tracking should measure instead of one blended position.
Does Reddit still dominate AI search citations?
Reddit was the most-cited domain on Perplexity and ChatGPT Search in late 2025. In August 2026 its ChatGPT citation share fell 86% in a single day, and nobody announced it.
Do reviews get you cited by AI search?
One September 2025 study ties Facebook and Yelp presence to businesses AI cites locally. A 2026 consumer survey shows the real thresholds: 20 reviews, 4.5 stars, a reply within a week.
Does keyword cannibalization actually hurt rankings?
A 2026 study of 100 sites found successful pages already average 4.7 competing URLs per keyword. Where it actually costs you: sites under DR 50, stuck at positions 8 to 20.
Where AI engines actually quote from on a page
Across 8,000 AI citations, 82.5% point to a specific page, not a homepage. Inside the page, 44.2% of the quoted text comes from the first third.
Do author bios actually get you cited by AI search?
A real test found bylines roughly doubled Bing citations and lifted crawl rate. Google's AI Overviews did not move, and unknown authors gained the most.
Core Web Vitals: real for Google, measured at zero for AI citations
The 2025 Web Almanac found only 48% of mobile origins pass all three thresholds. A 2 million-citation study found no measured effect on which pages AI engines actually cite.
How to get cited by Gemini
Ahrefs found the Gemini app matches Google's own top 10 only 8% of the time, versus 76% for AI Overviews built on the same models. What ranking has to do with citation, and what does not.
Does internal linking still matter?
A 23-million-link study found roughly 4x the clicks at 40 internal links versus under 5. Google confirms the mechanism. Nobody has measured the AI citation side yet.
Does duplicate content still matter?
Google's had no duplicate content penalty for 17 years. A 2021 study found deduplicating LLM training data cut memorization tenfold. What that means for canonical tags and AI citations.
Website migration recovery: what 1,052 migrations show
A 1,052-migration study found only 23% recovered traffic within 90 days, despite a headline claiming 27%. Median recovery: 304 days.
How to get cited by Claude
One tracked study found Claude cites about 3.6 sources per answer, the fewest of any engine, and mentions brands far more than it actually credits them.
Does journalist outreach still earn links?
HARO shut down in December 2024. A rival bought and revived it four months later. What the pitch data and new AI-citation studies say about pitching journalists today.
Do zero-volume keywords actually drive traffic?
Ahrefs tested its own advice on its own blog. Reported-zero keywords averaged 11.3 impressions a month, matching the estimate instead of beating it.
Does content pruning work?
A 14,000-URL case study turned a five-year decline into a 23% YoY gain, mostly without deleting anything. Google's Mueller drew a narrower, newer line in 2026.
Does crawl budget still matter?
Google's own crawl team says most sites can ignore it below a million URLs. Vercel's data says GPTBot and Anthropic's crawler still waste a third of every visit on dead ends.
Does hreflang still matter?
Ahrefs found 67% of 374,756 domains have a hreflang error. A December 2025 test found ChatGPT, Perplexity and Claude ignoring the signal that Google and Bing still read correctly.
Hiding your address on Google Business Profile
One tester found zero ranking impact from hiding it. Another watched a local pack disappear for a month. The difference was what the listing looked like before anyone touched it.
Does AI-generated content rank worse?
Ahrefs studied 331,000 pages, Graphite studied 31,000 keywords. Neither found a penalty for being AI-written, only for being low-effort.
Why your rank tracker shows a different number than what you see
Google expanded Personal Intelligence to nearly 200 countries in May 2026. Two studies on search personalization, 13 years apart, land in different places.
What is SEO share of voice, and can you trust the number
A Semrush worked example shows share of voice swinging from 8% to 4% between cities while rank barely moved. The real gaps in how it is calculated.
Does Domain Authority still matter?
Google has denied using it since 2015. An independent study of 2,000 keyword SERPs still found it correlates with rankings at 0.16, weak but not zero.
Does keyword density still matter?
A 2026 study of 1,536 Google results across 32 keywords found density holding flat near zero at every ranking position. Google calls repeating keywords a spam violation, not a formula.
Does social media still matter for SEO?
Google has said since 2013 that shares and likes are not a ranking factor. A 100,000-keyword study found Google AI Overviews citing a social platform in 68% of answers anyway.
How to read this section
These guides exist because we got tired of answering the same questions during scoping. They are ordered the way the work actually happens, not the way that makes the topic list look impressive.
Starting from zero, read what AEO is first, then GEO, AEO and SEO to settle the vocabulary, and stop there until something breaks.
Already ranking and still never cited? The cause is usually one of three things, and they have a diagnostic order. Access first: the AI crawlers guide tells you whether the engines can read you at all, and more sites fail this than expect to. Rendering sits underneath access: the AI crawlers do not run JavaScript, so a client-side page comes back empty even when every bot is allowed in. Identity second: entity SEO covers whether the engines are certain who you are, which is the quiet reason a smaller competitor gets named instead of you. Structure third: schema markup and AI Overviews cover how a machine lifts an answer off a page.
The llms.txt guide and getting cited by ChatGPT are the hands-on ones. Read those when you are ready to change something today. The Perplexity guide is the one engine-specific piece here, included because Perplexity documents more of its own behaviour than the others and is the only engine where you can check the result in your own referral data. Measuring AI visibility is the counterpart to all of them: what each instrument can prove, and what no dashboard can.
What we will not do here
Sell you a trend. llms.txt is a community proposal rather than a standard, and its own page says so in the first hundred words. AEO extends SEO, it does not replace it. Where the evidence is thin, these guides say the evidence is thin, because the alternative is you finding out later and trusting nothing else we wrote.
Every guide names the task that does the work, if you would rather buy it than run it yourself. That is the trade: real method in public, with a price next to it.