I've read a lot of "AI SEO" and "GEO" service pages this year while sizing up competitors. Most of them are the same 2023 deliverable list, keyword research, backlinks, a monthly blog post quota, with "AI" swapped in for "Google" in the headline. The audit template didn't change. The pricing tiers didn't change. Only the label did.
That gap matters because the underlying work actually did change. I covered the real shift in how to dominate SEO in the age of super intelligence, schema that validates, entity consistency, answer-first structure, and a product feed agents can read. None of that is a rebrand. It's new infrastructure. Here's the gap I keep seeing, and how to tell the difference before you sign a contract.
What "2023 SEO with a new label" actually looks like
I'm not naming names. This pattern shows up across dozens of small and mid-size agencies, not one bad actor. The tell is consistent: the proposal mentions ChatGPT and AI Overviews in the first paragraph, then the actual scope of work is identical to what the same shop sold in 2022.
| Deliverable | Relabeled 2023 SEO | Real AI-era SEO |
|---|---|---|
| Keyword research | Volume and difficulty spreadsheet, same tool, same output | Same research, plus a pass for question-shaped, answer-first query intent |
| Technical audit | Page speed, broken links, title tags | Same checks, plus schema validation (does it actually parse) and entity consistency across the site, socials, and mentions |
| Content | Monthly blog post quota, keyword-stuffed intro paragraphs | Answer in the first two sentences, FAQ sections with specific answers, internal links into a topic cluster |
| Reporting | Keyword rank tracker, organic sessions | Rank tracker, plus AI referral sessions as a separate GA4 channel, AI Overview impressions from Search Console |
| Commerce readiness | Not mentioned | Schema.org Product markup and Merchant Center feed audited, since shopping agents select off that feed directly |
The left column isn't useless. Page speed and broken links still matter. The problem is selling it as the answer to a question the client is actually asking, "will AI search find and trust my site," when nothing in the scope touches that question at all.
Ask any agency pitching "AI SEO" or "GEO" one question: "show me a page you've shipped where you added FAQPage schema, verified it parses in the Rich Results Test, and can show me the AI referral traffic in GA4 afterward." If the answer is a general explanation of what AI Overviews are instead of an actual page, you're looking at a relabel, not a rebuild.
Why the gap exists
Three reasons, and none of them are malicious. First, schema and entity work is genuinely more tedious than writing blog posts. It's markup, not prose, and it fails silently if you get the nesting wrong. Second, most rank trackers and reporting dashboards built for classic SEO don't separate AI referral traffic from organic search by default, so agencies report what their tools already measure, not what the client actually asked about.
Third, and this is the honest one, demand moved faster than supply. Clients started asking for "AI SEO" the moment ChatGPT and AI Overviews became part of daily search behavior. Agencies had a sales deck ready in weeks. Building the actual skill set, schema that validates, GTIN-complete product feeds, a tracking setup that isolates AI referral sessions, takes longer than updating a slide. I saw the same pattern play out at a terminology level when the White House renamed AI to "Super Intelligence" in the executive order I broke down here, the label changed overnight, the actual search behavior and ranking signals didn't move until months later, if at all.
There's a fourth reason I don't hear said out loud as often: a relabeled scope is easier to price. A monthly blog quota and a rank tracker have a known cost, agencies have sold that package for years and know their margin on it cold. Schema validation and entity cleanup are one-time, front-loaded work that doesn't map cleanly onto a recurring retainer, so it gets left out of the pitch even when the team knows how to do it.
What agent-readiness actually requires
I run the same structural checklist on every page across my own sites before it ships. None of it is exotic, all of it is skippable if you're just rebranding a keyword report. The common thread across all five items is that each one produces something you can check objectively, a schema validator either passes or fails, a GA4 channel either separates AI referral sessions or it doesn't. That's the opposite of a relabeled report, where the only evidence is a slide with a new logo on it.
- Schema that validates, not schema that exists. Article, FAQPage, Product, Organization, and Person schema, cross-referenced with consistent
sameAsentries, checked against the Rich Results Test. Copy-pasted markup that was never validated is often worse than no markup, it signals a site nobody maintains. - Entity consistency across the whole footprint. AI systems resolve "who is saying this" before deciding whether to trust it. A name that's slightly different on the site, LinkedIn, and old press mentions reads as noise, not as three mentions of the same source.
- Answer-first structure, measured, not assumed. A direct answer in the first two sentences under a question-shaped heading. I've tested this across affiliate and service pages, it's the single change I'd make first if I only had time for one.
- Tracking that actually separates AI referral traffic. A custom GA4 channel group matched on chatgpt.com, perplexity.ai, gemini.google.com, plus the Search Console Generative AI performance report. Without this split, you can't tell a client whether any of the work moved anything.
- Product feed readiness, if there's anything to sell. Complete Schema.org Product markup and a clean Merchant Center feed, since shopping agents inside ChatGPT, Gemini, and Perplexity select off that feed before a human ever lands on the page.
I wrote the full version of this list, with the exact JSON-LD and tracking setup, in the pre-launch SEO/AIO/GEO/AEO/SXO/LLMO checklist. None of it requires a new vendor relationship. It requires someone willing to open the Rich Results Test and actually read the error.
1. Show me one live page with FAQPage or Product schema you shipped,
and confirm it passes the Rich Results Test right now.
2. Show me your GA4 setup that separates AI referral sessions
(chatgpt.com, perplexity.ai, gemini.google.com) from organic search.
3. If I sell anything, what's your plan for my Merchant Center feed
and Schema.org Product markup, not just my blog content?
4. What does "AI SEO" change about my technical audit, specifically,
beyond the name of the report? Key takeaways
- A lot of "AI SEO" and "GEO" positioning is a relabeled 2023 scope of work, same deliverables, new headline.
- The tell is a proposal that mentions AI search in the pitch but never touches schema validation, entity consistency, or AI referral tracking in the actual scope.
- Real agent-readiness work is tedious and skippable by design, that's exactly why it's the differentiator right now.
- Schema.org Product markup isn't an SEO nicety anymore if you sell anything, shopping agents select off it directly.
- Ask for one live, validated page and one AI-referral GA4 view before signing anything. Either exists or it doesn't.
What I'll do next
I'm turning the four questions above into a one-page checklist I hand to anyone who asks me to vet an agency pitch before they sign, since I've now been asked three separate times this quarter. I'm also revisiting my own SEO in the age of super intelligence post to add a direct link to this one, so the critique and the checklist sit next to each other instead of in two separate posts.
FAQ
Is "AI SEO" or "GEO" a real discipline, or is it all relabeling? +
How do I tell if an agency's AI SEO service is real before I pay for it?
Does my technical SEO audit need to change for AI search?
Why do most agencies struggle to deliver real GEO work?
What's the fastest way to check if my own site is agent-ready?
Want this built for your team?
I design AI agents and growth automation that run without babysitting. If that sounds useful, let's talk.
Get in touch โ