The two senses, kept apart
Sense one is a visibility problem. Buyers increasingly start in an assistant, and the question is whether your brand is named and cited when they do. That work is mostly crawlability, evidence density and entity clarity, and it overlaps almost completely with what is published as generative and answer engine optimisation.
Sense two is a production problem. It asks whether AI can compress the research, drafting and analysis that an SEO programme requires. That is a question about cost and speed, and it has no direct effect on whether anyone is cited anywhere.
An agency selling "AI SEO" may mean either. The useful clarifying question is simple: does this change what my buyers see, or does it change how fast you work? Both are legitimate; charging for one while delivering the other is not.
What is genuinely new, and what is a rebrand
| Area | Genuinely new? | What it involves |
|---|---|---|
| Crawler permissions | New | Assistant operators publish distinct agents for training, search and user-triggered fetches, and a blanket block removes you from answers while leaving Google Search untouched OpenAI, bots and crawler documentation |
| Citation measurement | New | No operator publishes a per-site citation report, so measurement means a fixed prompt panel run on a schedule |
| Evidence density | Partly new | Adding quotations and cited sources measurably raised visibility in generated answers Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024. Good practice before; now it has a measured payoff |
| Entity consistency | Partly new | One description, one organisation node, one named person. Always mattered, matters more when a system has to resolve who you are |
| Crawlability and rendering | Not new | Content behind client-side JavaScript risks being seen late or not at all Google Search Central, JavaScript SEO basics. The same problem as a decade ago, with less forgiving consumers |
| Structured data | Not new | Describes visible content accurately Google Search Central, intro to structured data markup. No new schema type makes a page citable |
| Intent-first content | Not new | Writing for the question a buyer actually asks has not changed. It is simply harder to fake now |
The production sense, and where it goes wrong
Used properly, AI compresses the volume work: reading competitor pages at scale, extracting what each covers, producing a structured first pass. Used improperly, it produces unreviewed pages at volume, which is the failure mode Google addresses as scaled content abuse, judged by purpose and quality rather than by whether a human or a machine typed it Google Search Central, spam policies for Google web search.
There is also a commercial reason the review stage survives: unreviewed output is unsourced, and unsourced content underperforms in exactly the surface it is usually produced for. Adding cited sources was the single largest measured gain in the published GEO study Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024, and a model does not add a source it was never given.
Where the money should go first
- If assistant agents cannot reach your pages, fix that before anything else. It is a robots.txt and rendering problem and it is usually cheap OpenAI, bots and crawler documentation.
- If your pages contain no specific, sourced claims, fix that next. It is an editorial problem, not a technical one.
- If your entity is described five different ways across the site, fix that third. It costs a day and removes a whole class of ambiguity.
- Only then consider a separate AI SEO engagement, and ask what it does that the first three did not already cover.
Frequently asked questions
Is AI SEO different from SEO?
Partly. The crawler permissions, citation measurement and entity resolution work is new. Crawlability, rendering, structured data and intent-first content are the same disciplines as before.
Does AI SEO replace traditional SEO?
No. Ranking still drives most discovery for most businesses, and the technical work that makes a page rankable is the same work that makes it citable.
Is using AI to write content against Google guidelines?
No. Google targets scaled content abuse, meaning content made primarily to manipulate rankings rather than help people, whichever way it was produced Google Search Central, spam policies for Google web search.
How do I know if an agency is selling a rebrand?
Ask which of the three new areas they will work on, and how they will measure a citation. An answer that is only about writing more content faster is the production sense wearing the visibility sense’s name.
Can I measure AI SEO in Google Analytics?
Only partially. Assistant referrals undercount because buyers often search the brand afterwards and arrive as organic or direct.
What should a small team do first?
Check that the documented assistant agents are not blocked, then add sources to the claims already on the site. Both are cheap and both are prerequisites for everything else.