SEO vs AEO: The Metrics That Mattered, and the Ones That Matter Now

Classic SEO had a mature scoreboard: rank, share of voice, backlinks, traffic. Answer Engine Optimization is rewriting every line of it. Here is the old playbook next to the new one.

We have spent fifteen years doing classic SEO, the real kind, for some of the largest brands on the web. So we want to be precise about what is happening right now, because the loud version of this conversation is wrong in both directions.

SEO is not dead. AEO is not a buzzword. What actually happened is narrower and more interesting. The web grew a second discovery layer, and that layer needs its own scoreboard. The fundamentals that made classic SEO work, crawlable pages, fast load, clean structure, genuine authority, all still matter, because answer engines sit on top of search infrastructure rather than replacing it. Google's John Mueller said it directly: there is no GEO or AEO without SEO fundamentals, as Search Engine Land reported.

But the metrics, the things you actually put on a dashboard and report on, those have changed. A rank tracker cannot see a ChatGPT answer. Your backlink tool cannot count a mention that carries no link. If you are still measuring only the old scoreboard, you are flying half blind. So let us go line by line.

The old scoreboard and the new one, side by side

Now the detail, because each of those rows hides a real shift in how you work.

Keyword rank tracking becomes prompt coverage

For twenty years, the heartbeat of SEO was the rank tracker. You picked a universe of target keywords and watched your position, one through one hundred, by location and by device. Position was a clean, continuous number. Position three was meaningfully better than position eight, and you could chart it every single day.

Answer engines do not have a position three. A chat answer is a paragraph, not a list. The question is no longer where you rank. It is whether you are in the answer at all, and if so, how prominently, and with a link or just a name.

So the unit of tracking changes from the keyword to the prompt. You assemble a set of priority prompts, the real questions a buyer would actually ask an AI, phrased conversationally and at length, and you measure three things. Presence, whether you appear at all. Prominence, whether you are the lead recommendation or a footnote. Citation, whether the model links to you or merely names you. Coverage across that prompt set, the percentage of priority prompts where you show up, is the closest thing AEO has to a rank.

One hard truth the old world did not have to deal with: answers are not deterministic. Ask the same model the same question twice and you can get two different answers. A single check means nothing. You sample, repeatedly, and you watch the trend.

Share of voice becomes AI share of voice

Share of voice was always the metric we trusted most in classic SEO, because it was competitive. It did not just ask how you were doing. It asked how you were doing against the specific set of rivals fighting for the same keyword universe, weighted by search volume and position.

That instinct translates perfectly, and AI share of voice is arguably the single most important AEO metric. It asks: across your priority prompt set, how often does the model name you, compared to how often it names each competitor. If a buyer asks an answer engine to recommend a studio like yours, and your three rivals get named in eight of ten answers while you get named in two, that is the new rankings report, and it is brutal in its clarity.

The discipline is the same as it ever was. Define the competitor set honestly. Define the prompt set around real buyer intent. Sample on a schedule. Chart the trend. The substrate is new, the muscle is old.

Backlinks become mentions and live citations

This is the row with the biggest conceptual shift, so sit with it.

Classic SEO ran on the link graph. A backlink from a trusted domain was a vote, and authority flowed through hyperlinks. You built links, you measured them by the authority of the linking domain, and the link tool was sacred.

Answer engines do not run on the link graph in the same way. A large language model is shaped by the corpus it learned from, and for retrieval based answers, by the documents it pulls in live. In that world, an unlinked mention still counts. If a trusted industry publication describes your company accurately, the model can absorb that whether or not the words are wrapped in an anchor tag. The link graph becomes, in effect, a mention graph.

Two things follow. First, consistency is now an authority signal. When many credible sources describe you the same way, the model grows confident about who you are, and confidence earns citations. When sources disagree, the model hedges, and hedging keeps you out of the answer. Second, for retrieval engines like Perplexity and Google's AI Overviews, being cited as a source in a live answer is the true descendant of the ranking backlink, and it is worth tracking by name.

What earns those mentions and citations is not a mystery. The foundational study from Princeton and IIT Delhi tested it across thousands of queries:

Citing authoritative sources improved visibility for previously low ranked content by roughly 115 percent. Adding relevant statistics lifted it by about 33 percent. Including direct quotations raised it by about 43 percent.

Source: GEO: Generative Engine Optimization, Princeton and IIT Delhi, KDD 2024

The page that states something concrete and sources it is the page that gets pulled into an answer. Substance is the new link building.

Organic traffic becomes AI referral traffic, plus a blind spot

Classic SEO ended at a measurable click. Someone saw your result, clicked, and your analytics recorded the session. Click through rate told you how compelling your listing was.

AEO breaks that clean loop in two places. First, many AI answers are zero click. The user gets what they need inside the chat and never visits anyone. The influence was real, the session was not. Second, the traffic that does arrive shows up as AI referral traffic, humans landing on your site with a referrer of chatgpt.com, claude.ai, or perplexity.ai, and that traffic tends to convert well, because the visitor arrived pre sold by a recommendation.

Now the honest part. There is a blind spot. There is no Search Console for answer engines yet. You cannot see every answer you appeared in, the way you once saw every search impression. Search Engine Land has been clear eyed about this, arguing that LLM optimization will only mature once visibility becomes measurable, and that the industry is still in a pre Moz, pre Ahrefs era for AI. Measure what you can, and do not pretend the blind spot is not there.

So how do you actually measure the new scoreboard

Two pieces of instrumentation get you most of the way, and neither is exotic.

Capture the traffic the old tools miss. AI crawlers do not run JavaScript, so analytics never sees them. You read your server access logs instead, and bucket every request:

// Turn raw access logs into AEO traffic signal
const AI_CRAWLERS  = /gptbot|claudebot|perplexitybot|google-extended|oai-searchbot/;
const AI_REFERRERS = /chatgpt\.com|claude\.ai|perplexity\.ai|gemini\.google\.com/;

function bucketRequest(req) {
  const ua  = (req.userAgent || '').toLowerCase();
  const ref = (req.referer   || '').toLowerCase();

  if (AI_CRAWLERS.test(ua))   return 'ai_crawl';     // a model read your page
  if (AI_REFERRERS.test(ref)) return 'ai_referral';  // a human came from an AI answer
  return 'other';
}

Tally those two buckets weekly. The crawl count tells you the models can reach you. The referral count tells you humans are acting on what the models say.

Measure AI share of voice directly. Since no tool reliably reports this yet, you run the prompts yourself. The pattern is a list of priority prompts, a list of brands including your competitors, and multiple runs per prompt to average out the randomness:

// AI share of voice: how often each brand is named across
// your priority prompts. Run each prompt several times,
// because model answers are not deterministic.

const PROMPTS = [
  'best product development studio for a non technical founder',
  'who offers fractional CTO services in New York',
  'top agencies to build a SaaS MVP'
];
const BRANDS = ['Cause of a Kind', 'Competitor One', 'Competitor Two'];
const RUNS_PER_PROMPT = 5;

async function measureShareOfVoice() {
  const mentions = Object.fromEntries(BRANDS.map(b => [b, 0]));
  let answers = 0;

  for (const prompt of PROMPTS) {
    for (let run = 0; run < RUNS_PER_PROMPT; run++) {
      const answer = (await askModel(prompt)).toLowerCase(); // your model API call
      answers++;
      for (const brand of BRANDS) {
        if (answer.includes(brand.toLowerCase())) mentions[brand]++;
      }
    }
  }

  return BRANDS.map(brand => ({
    brand,
    coverage: (mentions[brand] / answers * 100).toFixed(1) + '%'
  }));
}

Run that on a schedule, store every result with a date, and you have rebuilt rank tracking and share of voice for the answer engine era. Search Engine Land lays out the same manual sampling approach in its guide to measuring brand visibility in AI search, and it is the right instinct while the tooling catches up.

The COAK take

Here is the reassuring part, and we mean it after fifteen years in this work. The disciplines rhyme. Prompt coverage is rank tracking. AI share of voice is share of voice. Mentions are the new backlinks. The questions a serious marketer asks, am I visible, am I winning against my competitors, am I trusted by credible sources, are exactly the same questions. Only the instruments changed.

So do not throw away the SEO playbook, and do not let anyone tell you the AEO one is mystical. Keep your fundamentals strong, because answer engines are built on them. Then add the second scoreboard: track your prompts, measure your AI share of voice, watch your referral logs, and earn consistent mentions across the sources the models trust.

For the technical foundation underneath all of this, see our companion piece Is Your Website Optimized For LLMs?, and our top three moves for becoming more visible in AEO. This article is the scoreboard. Those are the field manual.

If you want a partner who has run both playbooks, the classic one for the largest brands on the web and the new one for startups being born today, that is what we do. Cause of a Kind is full stack, full service, on shore and in house. We help cool people build great products, and we make sure both the search engines and the answer engines know their name.