LLM SEO? What Will They Think of Next?

Last updated: September 2026.

By Ryan Goering, founder and CEO of BaaDigi, a contractor marketing agency.

Why does the name keep changing?

Because three groups named the same thing. The 2024 Princeton paper that first measured it called it generative engine optimization (GEO). Marketing vendors call it answer engine optimization (AEO). People typing into a search box call it LLM SEO, LLM optimization or ChatGPT SEO. The labels differ; the job does not. If a proposal uses one of these terms, ask what the vendor will change on your pages and how they will measure it per engine. If the answer is a monthly 'AI visibility score' with no engine breakdown, the label is doing the work. Our comparison of AEO vs SEO for contractors covers how to read those proposals.

What did our own tracker show?

Between 18 and 31 August 2026 we ran 88 contractor-marketing questions through ChatGPT, Perplexity and Google AI Overviews three times a week and logged every cited domain. That produced 1,429 answers (511 ChatGPT, 511 Perplexity, 407 AI Overviews after failed runs were excluded). The full method and table are in the 1,429-search study. Three findings drive everything else on this page:

• The engines are not interchangeable. LinkedIn: 246 Perplexity citations, 6 ChatGPT. YouTube: 214 AI Overview citations, 4 ChatGPT. Advice tuned to one engine can be invisible in another.
• First-hand beats authoritative. Reddit was the most-cited domain at 657 citations, ahead of every marketing agency and publisher in the set. A contractor describing what they paid is better evidence than an agency describing what contractors typically pay.
• Google cites Google. google.com appeared 310 times, all inside AI Overviews. When Google has its own documentation on a question, you are competing with it, not replacing it.

Our own domain is in that table too (382 citations), partly because several prompts named us. Read that row as an artefact of the method, not a result.

What changes on a contractor page?

Four things, and none of them require a new website.

Put the answer first

Every page that answers a question opens with the answer in the first 40 to 70 words, before the story. The Princeton study measured this class of change directly: adding cited sources raised visibility in generative answers by about 40%, adding statistics by about 37% and adding quotations by about 30%, with the largest gains on lower-ranked sites. A model has to be able to lift a self-contained passage; if your answer is spread over six paragraphs, it lifts a competitor's.

Publish the numbers you actually have

Prices, service areas, response times, review counts, what a job included. A page with a real figure becomes a row in the model's comparison table; a page that says 'contact us for a quote' is skipped. We apply this to ourselves: the Stability Engine is published at $497 per month plus $995 setup, and our client data on speed to lead and Local Services Ads cost per lead is published with its sample sizes.

Make the facts machine-readable

Structured data for the business, the services and the FAQs, an accurate Business Profile, and an llms.txt file that tells a crawler where the definitive pages are. Our llms.txt generator builds one. Check that AI crawlers are not blocked in robots.txt; OpenAI documents its crawlers (GPTBot for training, OAI-SearchBot for ChatGPT search) and blocking one is not the same as blocking the other. Our guide to AI crawlers and robots.txt walks through the choice.

Be present where the engines already read

Reddit, LinkedIn and YouTube out-cited every agency in our sample. That is not a reason to spam them; it is a reason to make sure the honest, specific answers you give customers by phone also exist somewhere a model can read. Reviews on your Business Profile count here too: across 7,203 reviews on 37 client profiles, owners had replied to 51.5%, and the median reply took 90 hours. A reply is a first-hand statement in your own words on a page every engine crawls.

What stays exactly the same?

Crawlability, page speed, an accurate Business Profile, real reviews, internal links between your service pages and location pages. Across 1,047 free website audits run on our site between March and September 2026, 40.5% of sites had no schema markup, 27.1% had no XML sitemap, 18% failed the mobile-responsive check and only 57.9% had a Google Business Profile the audit could find. None of those sites need an LLM strategy yet; they need the foundation. Our home service SEO guide covers the foundation, and the SEO service page covers what we do about it.

How do you measure LLM SEO?

Per engine, never blended. Pick 10 to 20 questions a customer in your area would ask, run each through every engine that matters to you, record whether you are named and who is named ahead of you, and repeat weekly because answers vary between identical runs. Our free AI search visibility tool runs that check; the LLM SEO tool article explains what a paid tracker should and should not claim from the data, and how to check AI search visibility covers the manual version. For the baseline: across 254 businesses that ran the Perplexity check through our audit between June and September 2026, 59 (23.2%) were named for their own service-and-city question, with a median of three competitors named ahead of them.

Which engine matters for a contractor?

Google AI Overviews sit on top of the searches your customers already make, so for most local trades they matter first. ChatGPT matters when a customer researches a project before they search for a provider; see ChatGPT SEO for contractors for what gets a business named there and how to check whether ChatGPT recommends your business. Perplexity matters for research-heavy purchases and B2B work. Our tracker's per-engine split is the reason we report them separately in the dashboard rather than as one score; the write-up on which AI engine sends contractor leads goes deeper.

What does LLM SEO cost?

At BaaDigi it is not a separate line. Every page we publish for a client is written answer-first with published figures and structured data, because that is what gets quoted. That work is inside the Stability Engine at $497 per month plus $995 setup (website, Business Profile, content, reporting, and Google Ads management for up to $1,600 per month in ad spend paid to Google). Pricing for the larger engines is published on the Predictability Engine page. If a vendor quotes 'LLM SEO' as an add-on, ask which pages change and which engine the report will show.

Where to go next

• How do you get cited by AI? The page-level checklist.
• How to show up in AI search. The business-level checklist.
• We ran 1,429 AI searches. The data behind this guide.
• LLM SEO, LLM visibility and answer engine optimization in the glossary.