Loudachris
AI Search, Without The Sales Pitch

What Actually Works in AI Search

Most of what is sold as AEO, GEO or LLM SEO is either standard SEO with a new name on the invoice, or a tactic the documentation says is not needed. Below is the full list, each row linked to its source, then what genuinely changed and a method you can run yourself to check every claim on this page.

Chris Lourenco, Director of Loudachris Digital Marketing
Loudachris SEO specialist
Loudachris customer manager

Work directly with Chris and the team

Meet the team →
7 claims, 7 sourcesNo invented case studies · No promised timelines

The position in a sentence

AEO and GEO are SEO under new names. We sell a measurement diagnostic that tells you whether you have a problem, with the scope published, not a new discipline with a new budget line.

Changed: how answers are displayed
Changed: how you measure
Changed: how queries are phrased
Unchanged: what you actually do
Sourced → Testable → Yours

AI Overview

AEO and GEO are new names for SEO. What changed in AI search is presentation, measurement and phrasing: answers are summarised above the links, visits arrive with less usable referrer data, and people type longer questions. What did not change is the work. There is no llms.txt requirement, no AI-specific schema vocabulary, no chunking rule, and no way to guarantee a citation, because citation selection is probabilistic and varies by phrasing, session and model version. Loudachris publishes the scope of a free AI search diagnostic and folds any resulting work into standard SEO from $1,500+GST/month, with no separate AI fee.

The Teardown

Every tactic being sold, and what the documentation says.

These are the 7 items that turn up most often in AI search proposals. Each row links to the primary source rather than a summary of a summary, so you can read the wording yourself before you sign anything. We are not naming the firms selling these. The tactic is the problem, not the letterhead.

Sold asStatusSource
llms.txt fileGoogle's AI features guidance states it is not neededGoogle Search Central
AI-specific schema vocabularyNo such schema.org vocabulary exists for generative AIGoogle Search Central
Content chunking "for AI"No requirement to break content into small piecesGoogle Search Central
Writing specifically for AINot required. Write for peopleGoogle Search Central
Speakable schema as an AI leverLimited availability, news publishers onlyGoogle Search Central
HowTo rich resultsRetired from Google SearchGoogle Search Central
Guaranteed AI citationsCitation selection is probabilistic and varies by phrasing, session and model versionGoogle Search Central

Sources checked 9 August 2026. If any of them change, this table changes with them and the review date at the bottom of the page moves.

There is a fair reading of some of these. An llms.txt file costs nothing to publish, and clear section headings are good writing on their own merits. The objection is not that the tactics are dangerous. It is that they are priced as a new discipline, on a monthly retainer, with an implied outcome nobody can deliver. Charge only for what the documentation supports and the invoice looks very different.

What genuinely changed.

Something real did happen. It is just not the thing being sold. Here are the 3 shifts that survive scrutiny.

How answers are displayed

A generated summary now sits where a list of 10 blue links used to be. For a question with a factual answer, the reader can finish the task without clicking anything. For a question that ends in someone being hired, the click still has to happen, because a summary cannot quote a price for your job or answer the phone. That split, informational versus commercial, decides how much any of this matters to you.

How you measure

A visit that started inside an assistant often arrives without a referrer you can segment cleanly, and a generated summary does not report the way a ranking position does. The result is a blurrier picture, not a broken one. The fix is measurement discipline: track enquiries and calls per query family rather than chasing a single number, and be honest in reporting about which parts are estimates.

How queries are phrased

People type fragments into a search box and whole sentences into an assistant. A query that used to be 2 words now carries a suburb, a constraint and a budget in a single sentence. Pages that state plainly what you do, where you do it, what it involves and roughly what it costs answer more of that sentence. Same advice as always, with a sharper reason behind it.

The Unchanged Part

What did not change: the work itself.

Presentation moved. Measurement moved. Phrasing moved. The task list did not. Everything below was the job before any of this arrived and it is still the job now, which is the whole reason we do not sell a second retainer beside the first. Put plainly: this is SEO.

Being crawlable and indexed

Nothing can quote a page it cannot fetch, render or index. Blocked resources, broken canonicals and pages stuck outside the index are still the first thing to fix, and still the least glamorous item on any proposal.

Genuinely useful, specific content

A page that states what the service includes, which areas it covers, what it involves and roughly what it costs gives a summary something concrete to draw on. A page that says quality service, competitive prices gives it nothing.

Accurate, consistent business data

A single version of your name, address, phone number, hours and service list across your site, your Google Business Profile and the directories that mention you. Contradictions are the cheapest reason to be left out of an answer.

Real reviews from real customers

Volume, recency, and the actual words customers use. Reviews are the raw material any summary about a local business reaches for, and they are the signal a competitor cannot fabricate on your behalf.

Standard schema.org markup

Correctly implemented, honest markup describing the organisation, the services and the pages. Not because it is an AI lever, but because it has been part of competent technical SEO for years and is still worth doing.

Links and mentions from real sites

Earned coverage from sites that would exist whether or not you paid them. This has survived every algorithm reshuffle so far, and it survives this one too, because it is the hardest signal to manufacture.

Verify It Without Us

The 5-minute method: check any claim on this page yourself.

You should not take our word for this, and you should not take anyone else's either. This takes 5 minutes, needs no software, and produces a number you own rather than a screenshot somebody sent you.

1

Write the question the way a customer would say it

Not your keyword. Their sentence, with the suburb, the constraint and the urgency in it, because that is how people actually type to an assistant. Start with a single question.

2

Open a brand new session, signed out if the tool allows it

Chat history and account personalisation both change what comes back. A test run in a session that has already discussed your business is not a test, it is a mirror.

3

Ask it, and write down what you get

Record 2 things only: which businesses are named, and which URLs are cited. Ignore the prose. The named list and the source list are the parts that matter.

4

Repeat the identical prompt at least 3 times, fresh session each time

Same wording, new session. Then, if you have 60 seconds spare, run it once on a second assistant. You now have a small sample instead of an anecdote.

5

Count, do not conclude

Your result is a fraction: named in 2 of 4 runs, cited in 1 of 4. That fraction is the honest output. It is not a ranking, it has no trend yet, and it should be re-run before you spend money on the strength of it.

Now point the same method at a sales pitch

When a provider shows you a screenshot of an AI answer naming a business, ask 3 questions. How many times was the prompt run? Were the sessions fresh? And can you re-run the identical prompt right now, in front of them? A claim that survives all 3 is worth listening to. A claim that fails the first was never a measurement.

The Variance Principle

The same prompt, 3 times, can give 3 different answers.

This is the load-bearing idea on the page, and most of the dubious claims in this category collapse the moment you hold on to it. Generated answers are not looked up from a table. They are produced, and the production varies with how the question was phrased, what else was in the session, which model version is serving you, and what got retrieved at that moment.

So a binary claim is unsound in both directions. A provider who says you do not appear in AI answers, on the strength of a single check, has not measured anything. A provider who shows you an answer naming your business, on the strength of a single check, has not measured anything either. The unit of evidence here is a frequency across repeated runs, not a yes or a no.

A single check is not evidence

A single run tells you what happened in that run. Treat it the way you would treat stopping 1 person on a street corner to ask for the best cafe in town: interesting, not conclusive, and dangerous if you budget against it.

Frequency is the honest metric

Named in 3 of 10 runs is a statement you can defend, act on, and re-measure next quarter. Visible in AI search is not a statement at all. Insist on the first shape and most of the noise in this market disappears.

Comparisons need identical conditions

Before and after only means something when the prompt, the session state and the assistant are held constant, and when both sides used the same number of runs. Change any of those and you are comparing 2 different tests.

What we are deliberately not putting on this page

We have not published the results of a study we have not run. No percentages, no growth curves, no client outcomes attributed to AI visibility work. The method above is here so you can generate your own numbers, on your own business, and hold anyone who quotes you a figure to the same standard. If we ever publish numbers on this page, they will arrive with the prompt, the run count, the dates and the assistant used.

What we sell instead: a diagnostic, not a discipline.

If the tactics are hollow and the work is unchanged, the only honest product left is measurement: a read on whether you have a problem worth solving, run the way the method above describes, with the scope published before you ask for it. That audit is free. Whatever it turns up is ordinary SEO work, and it belongs inside an ordinary SEO retainer.

Free

The AI search diagnostic

Repeated prompts, recorded frequencies, and a plain read on whether any of this is costing you enquiries. The full audit scope is published, so you know what arrives.

From $1,500+GST/month

If there is work to do

Our published SEO pricing, month to month, no lock-in. The same figure whether or not the letters AI appear anywhere in the conversation.

None

The AI surcharge

There is no second line item, because there is no second discipline. If a quote in front of you carries one, ask which deliverable it buys that standard SEO does not already include.

The rest of this silo argues the same case from different angles: the AI search visibility hub, the case against answer engine optimisation and generative engine optimisation as separate products, what to ask an AI search agency, and how any of it applies here in Adelaide.

Read This Before You Spend Anything

Who does not need to think about this yet.

Most businesses asking us about AI search have a bigger, duller problem sitting in front of it. If any of the 5 below describes you, this page is interesting reading and nothing more. Fix the thing in the way first, because it will move revenue further and cost you less.

Your site is not properly indexed

If Search Console shows your key service pages sitting outside the index, no answer engine can reach them either. Indexing is a quality and crawl conversation, and it comes before every sentence on this page.

Your Google Business Profile is thin or wrong

Incomplete categories, missing service areas, an old phone number, a handful of ageing reviews. For a local business this is the highest-yield work available, and it happens to be the same data any summary about you would draw on.

Your enquiries come mostly from referrals and repeat customers

If search is not yet a meaningful channel for you, optimising a slice of search is optimising a slice of nothing. Build the channel first, then worry about how it gets presented.

The traffic you already have is not converting

Sending more visitors to a page that does not turn them into calls is an expensive way to make the same problem bigger. Conversion work is cheaper, faster, and fully within your control.

You cannot yet name your top 10 commercial queries

Without that list there is nothing to test, because the method above starts with the question a customer would ask. Get the list, run the method, then decide whether there is a problem at all.

None of that describes you? Run the method, get your fraction, and bring it to the free audit. We will run it independently and tell you whether the 2 sets of numbers agree.

Who wrote this

Chris Lourenco, Director, Loudachris Digital Marketing

Loudachris is a Google Partner agency at 21-22 Greenhill Road, Wayville SA 5034, working with Australian service businesses on SEO, Google Ads and the websites that support them. This page exists because we kept being asked to quote for AEO and GEO packages, and the honest answer was that we would be billing twice for work already inside the retainer. Everything here is sourced or testable, and anything we could not source has been left out rather than softened into a claim.

Disagree with something here, or found a source that has since changed? Send it through the contact page and we will correct the page and move the review date.

ABN 73 630 143 190Wayville SA 5034Last reviewed: 9 August 2026

Straight answers about LLM SEO and AI search.

Is LLM SEO a real discipline, or is it just SEO?
It is SEO. The label is new, the deliverables are not. Every input an AI answer can draw on was found by crawling the open web, reading your business data, or retrieving a page a search index already knew about. That is the same surface SEO has always worked on. The honest test for any pitch: ask which deliverable in the proposal would be dropped if AI assistants disappeared tomorrow. If the answer is none of them, you are being quoted for SEO under a second invoice line.
Does SEO still matter for AI search?
More than before, because AI answers are assembled from sources that have to be found first. A page that is blocked, slow to render, thin, duplicated across a dozen near-identical URLs, or absent from the index cannot be quoted by anything. The uncomfortable version of this: if a site is not earning organic visibility now, adding an AI layer on top does nothing, because there is nothing underneath to retrieve.
What actually changed, then?
Only 3 things, in practice. How answers are displayed, so a summary now sits where a list of links used to be. How you measure, because a visit that started inside an assistant does not always arrive with a referrer you can segment. And how queries are phrased, because people type whole sentences to an assistant and fragments to a search box. What did not change is the work: crawlable pages, accurate business data, real reviews, standard markup, and content worth quoting.
Do I need an llms.txt file?
No. Google's AI features guidance states it is not needed, and we will not bill you to add one. It costs nothing to create, which is exactly why it is such a convenient line item: it looks like a deliverable, it ships in minutes, and nobody can prove it hurt anything. That is not the same as it helping.
Is there a schema type that AI systems reward?
No. There is no schema.org vocabulary for generative AI. Standard markup that accurately describes your organisation, your services, your location and your pages is worth implementing on its own merits, and it has been part of competent technical SEO for years. Anyone selling markup as an AI exclusive is selling you something your SEO should already include.
Should I chunk my content into small blocks for AI retrieval?
There is no requirement to break content into small pieces. Clear headings, short paragraphs and a direct answer near the top are good writing, and they were good writing before assistants existed. The version we argue against is different: taking a page that reads well and shredding it into machine-shaped fragments a human then has to wade through. That trades a real audience for an imagined one.
Should I write pages for AI instead of for people?
No. Writing specifically for AI is not required. Write for people. The practical reason matters more than the principle: the person who clicks through from an AI answer is the one who decides whether to call you, and a page tuned for a machine reads like a specification sheet to them. Optimise for the human at the end of the journey and the machine in the middle gets what it needs anyway.
Can anyone guarantee my business gets cited in ChatGPT or an AI Overview?
No, and a guarantee is the clearest signal to walk away. Citation selection is probabilistic and varies by phrasing, session and model version. A provider who promises a citation is either promising something outside their control or planning to show you a screenshot from a single lucky run. Neither is evidence.
Why does the same question give me a different answer every time I ask it?
Because that is how these systems work. The same prompt run 3 times can return 3 different answers, with different businesses named and different sources cited. This is the most important thing to understand before you buy anything in this category, because it means a single check proves nothing in either direction. Appearing once is not proof of visibility. Missing once is not proof of a problem.
How do I check whether my business appears in AI answers?
Run the same question repeatedly rather than once, in a fresh session each time, and record how many of those runs name you. The 5-minute method on this page walks through it, and it needs no tools you do not already have. Treat the output as a frequency out of the number of runs you did, never as a yes or a no.
What does your AI search audit cost?
The audit is free and the scope is published, so you can see exactly what gets measured before you ask for it. See the audit scope. If it finds work worth doing, that work sits inside our normal SEO retainer from $1,500+GST/month, month to month with no lock-in. There is no separate AI line item, because there is no separate AI discipline to bill for.
Who should ignore AI search for now?
Anyone whose site is not indexed yet, anyone whose Google Business Profile is incomplete, anyone whose enquiries come almost entirely from repeat customers and word of mouth, and anyone who has not yet fixed conversion on the pages already receiving traffic. In every case there is a larger, cheaper problem sitting in front of this one, and solving it first will move revenue further.

Find out whether you actually have a problem.

Run the 5-minute method yourself, then have us run it independently and compare. The audit is free, the scope is published, and if the answer is that nothing needs doing, that is the answer you will get.

Call 0403 454 199

Start at the AI search visibility hub or read the SEO pricing guide.

CallBook Free Audit