GEO Is SEO With Direct Answers. A 30-Day Test With Real Numbers.

Some agency has already messaged you about GEO. Your site needs to show up inside ChatGPT and Perplexity answers, Google is dying, and they can handle it for a monthly retainer. One pitch I got forwarded quoted INR 55,000 a month for "AI visibility management".

Here is my claim: GEO is SEO with direct answers. The tactics that get a page cited inside an AI answer are the same tactics that have ranked pages on Google for a decade. Same inputs. Same files. Same crawlers, mostly. Which means there is no tradeoff to manage and nothing exotic to buy. You can test the whole thing yourself, free, in 30 days.

I'm running that exact test on webight.com right now. The full protocol is below. Copy it.

What a GEO retainer is actually selling

Strip the pitch deck and every GEO service I've seen sells four things.

Answer-first paragraphs. The first 2 or 3 sentences under a heading should directly answer the question the heading asks. Price, timeline, yes or no, the number. AI engines lift these blocks almost verbatim because they're cheap to extract. But Google has wanted this since featured snippets launched in 2014. A paragraph that answers "how much does a landing page cost" in its first sentence wins the snippet AND the citation. One edit, two channels.

Entity clarity. The engine has to know, without guessing, who you are, where you are, and what you do. "Webight is a 2-person design and development studio in Delhi NCR that builds websites on Next.js, starting at INR 7,999 for a landing page." That sentence is boring. It is also machine-readable gold, and it's the same Organization schema and consistent NAP data that local SEO consultants have charged for since forever.

FAQ coverage. Real questions, answered in full sentences, marked up with FAQPage schema. AI engines train and retrieve on question-answer pairs. Google has rendered FAQ rich results from the identical markup for years.

Freshness dates. A visible "last updated" date plus dateModified in your schema. Perplexity weights recency hard. So does Google for any query with freshness intent.

Four tactics. Every single one is classic on-page SEO wearing a new acronym. See, that's the entire trick: the agency renamed work you should already be doing and attached a retainer to the rename.

I already proved the overlap on my own site this month

In June I audited webight.com properly and found embarrassing stuff. Our canonicals pointed to webight.com while the server 307-redirects everything to www.webight.com. Two URLs claiming to be the same identity. Google hates that. Guess what, so does an AI engine trying to resolve "Webight" into one entity. The fix for the SEO problem and the fix for the entity problem were the same fix.

The sitemap still listed 12 deleted case-study URLs. Dead references make a crawler trust the rest of your sitemap less, whether that crawler feeds an index or a language model.

The hero headline had a 4,190ms LCP element render delay because a framer-motion opacity animation sat waiting for hydration before showing text. I rewrote it as a pure CSS animation and the render delay dropped to 382ms. I turned on experimental.inlineCss in next.config.ts, which removed both render-blocking stylesheets, 22.5 KiB gone. Killed two leftover third-party scripts. Added security headers. Mobile scores went from 90/96/96/100 to 96/100/100/100.

Why does page speed matter for GEO? Because the bots fetching your pages for AI answers have timeouts and budgets like any other bot, and a page that paints in 382ms gets fully read more reliably than one stalling past 4 seconds. Every fix on that list helped both channels at once. Zero fixes helped one and hurt the other. That's the pattern, and it held through the entire audit.

The 30-day test protocol

This is what I'm running on webight.com. Started June 3. It costs time, not money.

Day 0: build the prompt sheet. Write 27 prompts a real prospect would type into an AI engine. Mine include "best web development studio in Noida for startups", "how much does a Next.js website cost in India", "agency that builds WhatsApp booking bots for hotels". Run all 27 in ChatGPT, Perplexity, Google AI Mode, and Gemini. Log every response in a spreadsheet: cited or not, linked or not, what got said about you. Use fresh chats, logged out where possible. This is your baseline. Ours was rough: webight.com appeared in 2 of 27 prompts.

Days 1 to 5: rewrite for answer-first. Pick your 9 most important pages. For each H2 that asks an implicit question, make the first sentence the direct answer. Our pricing page used to open with positioning copy. Now it opens with "Landing pages start at INR 7,999 ($149). Full websites start at INR 24,999. 50% upfront, 50% on delivery, with a 7-day money-back window after kickoff." A human skimmer and a retrieval system want the identical thing here.

Days 5 to 8: entity and FAQ markup. One canonical domain, one consistent description of who you are across every page, Organization schema in the layout, FAQPage schema on pages where you genuinely answer questions. Here is the FAQ component we ship, complete:

// app/pricing/FaqSchema.tsx
const faqs = [
  {
    q: "How much does a landing page cost at Webight?",
    a: "Landing pages start at INR 7,999 ($149). Full websites start at INR 24,999. Payment is 50% upfront and 50% on delivery, with a 7-day money-back window after kickoff.",
  },
  {
    q: "What stack does Webight build with?",
    a: "Next.js, TypeScript, Supabase or Postgres, Tailwind, and Vercel, with n8n and Resend for automation and email.",
  },
  {
    q: "Where is Webight based and who do you work with?",
    a: "Webight is a 2-person studio in Delhi NCR, India. Clients are in India, the Gulf, the US, Southeast Asia, and the Netherlands.",
  },
];

export default function FaqSchema() {
  const schema = {
    "@context": "https://schema.org",
    "@type": "FAQPage",
    mainEntity: faqs.map((f) => ({
      "@type": "Question",
      name: f.q,
      acceptedAnswer: { "@type": "Answer", text: f.a },
    })),
  };
  return (
    <script
      type="application/ld+json"
      dangerouslySetInnerHTML={{ __html: JSON.stringify(schema) }}
    />
  );
}

Drop it into the page, render the same questions visibly in the UI, done. The markup that earns Google rich results is the markup that hands an AI engine clean question-answer pairs. One component, two channels, again.

Days 8 to 30: freshness and weekly logging. Add visible updated dates and dateModified to anything you touched. Then rerun the full 27-prompt sheet every 7 days and log it in the same spreadsheet. Don't cherry-pick. Log the misses too, because the misses tell you which pages still bury their answers.

That's it. That's the whole protocol. We're 7 days in as I write this, and Perplexity has already started citing the pricing page in 3 prompts where it was invisible at baseline. Small. Early. I'll publish the full sheet when the 30 days close, hits and misses both, because a test you only report when it flatters you is marketing, and I'm trying to do the opposite here.

What this means for the retainer pitch

If GEO needed different work than SEO, you'd face a real decision and maybe a real vendor. It doesn't. Every tactic above is one edit serving both channels, which means a separate GEO retainer is paying a second team to redo your on-page work under a new name. There's a psychology principle underneath these pitches: new acronyms feel like new risk, and people pay to make unfamiliar risk go away. That fear is the product. The work was never new.

If an agency shows you their own prompt sheet, their own baseline, and their own before-and-after citations, fine, in that case they're doing the test I just described and charging for the labor. That can be worth it if you have zero time. But ask for the spreadsheet before you sign anything. No spreadsheet, no retainer.

So before you pay anyone INR 55,000 a month to say "GEO" at you, spend one evening writing 27 prompts and logging where you stand. Rewrite your top pages answer-first. Ship the schema above. Rerun the sheet in 30 days. If the numbers move, you did GEO yourself for free. If they don't, you just saved a retainer that wouldn't have moved them either.

DMs are open if you run it. Send me your sheet, I want to see the numbers.