What an LLM SEO consultant actually does — and how it differs from SEO

An LLM SEO consultant works to get your pages quoted inside AI answers rather than ranked beneath them. Across 7,108 polls of ChatGPT, Gemini and DuckDuckGo on 132 buyer prompts between 19 May and 13 September 2026, classic search presence preceded every ChatGPT citation we earned, and only 27% of prompts where we had it earned one.

At a glance

• The definition: becoming the source a large language model quotes when someone asks it a buying question, measured as citation share on a fixed prompt set, per engine, over time.
• The boundary: not a replacement for SEO, not a file at your site root, not a rank tracker with a new label.
• The labels: GEO has an academic origin (a November 2023 paper, later at KDD 2024). AEO, LLM SEO and AI SEO are market coinages with no standards body behind them.
• The measured split: Ahrefs found 37.9% of URLs cited in Google AI Overviews also ranked in the top ten blocks of that query’s SERP (863K SERPs, 4M URLs, March 2026), against roughly 76% in its July 2025 study.
• Google’s own position: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”

What does an LLM SEO consultant actually do?

Strip the label off and the job is four workstreams, only one of which a classic SEO retainer covers.

1. Build a prompt registry. Thirty to a hundred and fifty questions your buyers would actually type into an assistant, in their own words — “why is my sales conversion rate so low”, not “sales conversion rate improvement”. It replaces the keyword list, and a keyword is two words where a prompt is a sentence.
2. Poll and log. Ask each engine those prompts on a schedule and record, per poll, whether your domain was cited. Our rig polls ChatGPT and Gemini about twice a week and DuckDuckGo every day — DuckDuckGo ran on all 118 days of the window, ChatGPT and Gemini on 34 of them. The instrument is cheap; running it in the dull weeks is what most programmes skip.
3. Split the misses into two piles. Prompts where you are also absent from classic search are an SEO problem. Prompts where you rank and are still not quoted are an extractability problem: the answer is buried, the facts sit in prose, the page has no number anyone can lift.
4. Fix the entity layer. Consistent name, description and founding facts across your site, LinkedIn, Crunchbase and Wikidata, with sameAs pointing at all of them. Slowest lever, longest lag.

An LLM SEO consultant runs a measurement loop; everything else is a hypothesis waiting for the next poll. The operating cadence, week by week, is on our AI SEO and AEO service page.

What LLM SEO is not

Three claims are made constantly under this label. None survives a log file.

It is not a replacement for SEO. Of the 60 prompts we polled on both ChatGPT and DuckDuckGo in the window, our domain never once appeared on the DuckDuckGo answer page for 23 — and ChatGPT cited us on none of those 23. Classic search presence was a precondition, not an optional extra.

It is not a file you publish. Ahrefs checked 137,210 domains in May 2026 and found that of the roughly 38,000 with a valid llms.txt, 97% received no requests for it at all. Google’s documentation says you do not need machine-readable files, AI text files or markup to appear in its AI features, and adds that there is “no special schema.org structured data that you need to add”. The full evidence is in our writeup on whether llms.txt does anything for AI search.

It is not rank tracking renamed. A ranking is stable enough to screenshot monthly. A citation is not: in our published dataset, once ChatGPT first cited us on a prompt it cited us on a median of about 44% of that prompt’s polls from the first citation onward, against roughly 70% on Gemini. One screenshot of an AI answer, good or bad, is noise.

LLM SEO vs AEO vs GEO vs AI SEO — do the labels mean different things?

One of the four has a definition you can cite. GEO — generative engine optimisation — comes from an academic paper, GEO: Generative Engine Optimization by Aggarwal et al., first submitted on 16 November 2023 and presented at KDD 2024. It defines GEO as a black-box optimisation framework for improving content visibility in generative engine responses, and reports boosting visibility by up to 40%.

AEO (answer engine optimisation), LLM SEO and LLM optimisation are market coinages: no standards body ratified them, no paper defines them, and vendors use all three for the same work. AI SEO is the odd one out, because it describes how the work is produced — SEO delivered with AI tooling — rather than which surface it targets. The acronym tells you nothing about scope. Two consultants selling “LLM SEO” can be running entirely different programmes; the only thing separating them is whether either can show a poll series.

Which classic SEO signals move citation share, and which do not?

First-party rows come from our own citation monitor: 7,108 successful polls of ChatGPT (browse-enabled), Gemini (search-grounded) and DuckDuckGo answer pages against 132 buyer prompts, 19 May to 13 September 2026, one domain, one niche. The ChatGPT and Gemini cross-tabs below rest on the 60 and 68 prompts respectively that were polled on that engine and on DuckDuckGo inside the window.

Row one is the rule worth naming: the floor test — classic search presence is the floor of LLM SEO, not the answer. On ChatGPT it gated every citation we earned and still predicted only about one in four. Two caveats we would rather state than have you find: one domain in one niche, so treat it as directional, and the DuckDuckGo figure measures answer-page presence, a lower bar. The 37.9% and 76% also come from two Ahrefs studies with different methodologies, which is why we unpacked the ranking-to-citation decoupling separately.

Why “are we in AI search?” has three different answers

AI visibility is not one number, and the divergence is large enough to change what you buy. ChatGPT never cited us on a prompt where we were absent from the classic answer page; Gemini did so on 4 of 25 such prompts, because it grounds differently. Across the three engines, 74% of the prompts we were ever cited on were cited by exactly one of them. An LLM SEO consultant reporting a single “AI visibility score” is flattening away the only variable that tells you where the work should go — the breakdown sits in our writeup of 5,051 first-party AI citation polls.

When do I need LLM SEO work, and when is it the wrong spend?

The crossover between DIY and handing over is a prompt count and a cadence, not company size.

In the first three rows you do not need to hire anyone this quarter, and our guide to tracking your own AI search visibility is the method written down. The cost of the DIY route is not money, it is attention: a weekly loop only works if someone runs it every week.

The conditions that change the answer

• Your buyers use one engine heavily. Entity-graph work is the Google lever; classic search presence plus extractability is the ChatGPT lever. The two programmes share the on-page work and little else.
• The engines change underneath you. Ahrefs attributes part of the drop in top-10 sourcing to query fan-out — one question split into sub-queries, citing pages that surface across all of them. Every figure here has a shelf life; re-measure quarterly.
• You already publish original data. Extractability work pays fastest: your pages hold something nobody else has.

Frequently asked questions

Is LLM SEO just SEO with a new name?

Partly, and it is worth naming which parts. Crawlability, indexation, internal linking and the entity layer are the same work, and in our polling classic search presence gated every ChatGPT citation we earned. What is new is the unit of measurement: a prompt not a keyword, a citation not a position, a poll series not a rank check. Google’s AI features guidance says there are “no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary” — that is the honest floor. The ceiling differs: Ahrefs found only 37.9% of AI Overview citations came from top-10 URLs in March 2026, against roughly 76% in July 2025.

What is the difference between LLM SEO, AEO and GEO?

In vendor usage, nothing consistent. GEO is the only one with a formal origin: the 2023 paper GEO: Generative Engine Optimization, presented at KDD 2024, which frames it as optimising visibility inside generative engine responses. AEO and LLM SEO are market labels for the same activity. Judge a proposal on which engines it measures and how often, not on the acronym.

Do I need an LLM SEO agency rather than a consultant?

That is a delivery-model question, not a scope question: the scope is identical either way — a prompt registry, a poll series, extractability work and the entity layer. Before comparing structures, check whether either can show you a logged poll series for a client rather than a screenshot of one answer.

Does an LLM SEO consultant replace my SEO agency?

Not on this evidence. Of the 60 prompts we polled on both ChatGPT and DuckDuckGo, our domain never appeared on the classic answer page for 23, and ChatGPT cited us on none of those. The two overlap on technical and entity work, so running both without a clear split usually means paying twice for the same audit.

How would I know if LLM SEO is working?

By a citation rate on a fixed prompt set that moves, not by a screenshot. Baseline before any work starts, poll on a schedule, report per engine. In our data ChatGPT cited on a median of about 44% of a prompt’s polls from the first citation onward, so any single check sits inside the noise. The mechanics are in our guide to being in the 15% of pages ChatGPT actually cites.

Is there a certification or standard for LLM SEO?

No. There is no ratifying body, no agreed metric definition and no audited benchmark, which is why two suppliers can quote incompatible “AI visibility” numbers for the same domain. The only portable evidence is a dated, per-engine poll series with the prompt set disclosed.