The Answer Engine Is Not Your Intern: What Actually Works in SEO, AEO, and the Great llms.txt Gold Rush

For years, the web promised that it would reward the best information. Then it rewarded the best keyword density. Then the best backlink profile. Then the fastest site with the least offensive pop-up. Now, in the age of AI Overviews, answer engines, chat interfaces, and search results that seem increasingly determined to save us from clicking anything at all, everyone is asking a more anxious question: How do I become the answer?

The short answer is less glamorous than the consultants would prefer. Make the best, clearest, most original page for a real person who has a real reason to need it. Make it technically accessible. Give it a point of view rooted in experience, proof, and specificity. Then make peace with the fact that nobody—not Google, not ChatGPT, not the person selling a $4,000 “GEO visibility sprint”—can guarantee that a machine will cite you.

The longer answer is that SEO and AEO—Answer Engine Optimization, the new umbrella term for trying to appear in AI-generated responses—are not separate empires. They are neighboring countries sharing an increasingly porous border. The durable tactics remain surprisingly old-fashioned. The fashionable tactics range from mildly useful to pure séance.

The Fundamental Things Still Apply

There is an unfortunate tendency to discuss AI search as if a search engine has recently developed a soul. It has not. It has developed a new interface. Beneath the elegant summary, the confident paragraph, and the unnervingly plausible recommendation, there is still a system trying to retrieve, assess, organize, and present information from the web.

That means the basics are not “basic” in the dismissive sense. They are the work.

Google’s own guidance for its AI search features is almost comically unromantic: create helpful, reliable, people-first content; make pages crawlable; use conventional technical SEO; and provide material that is genuinely useful rather than content engineered merely to manipulate rankings. Google also says that its generative search features do not require special AI markup, new machine-readable files, or an llms.txt file

For a local business, a creative agency, a production company, a law firm, a SaaS product, or a specialist service provider, the page most likely to survive the transition from ten blue links to one synthetic answer is not a page that says, “We are passionate about excellence.” It is a page that answers the questions a customer would actually ask before spending money.

Consider the difference.

A weak videography-services page says:

“We provide high-quality video production solutions for brands of all sizes.”

A useful one says:

“Atlanta businesses hire us for commercial video, customer testimonials, event recap films, and social cutdowns. A one-day testimonial shoot typically includes a producer, a two-camera interview setup, b-roll, lighting, audio, and edited versions for your website, YouTube, LinkedIn, and paid social.”

The first is business-card copy. The second contains the ingredients of an answer: who the service is for, what it includes, where it happens, what the customer receives, and how the work becomes useful afterward. It gives both a human being and a retrieval system something concrete to hold.

What fundamentally works is content that reduces uncertainty.

That includes:

Clear pages built around specific customer questions and decisions.
First-hand information unavailable in generic summaries.
Demonstrable expertise, especially when the subject involves money, risk, health, legality, technical implementation, or a major purchase.
Accurate product, service, pricing, location, availability, and policy information.
Original research, case studies, comparisons, process explanations, footage, photos, data, and examples.
Strong internal linking that makes the site understandable as a connected body of work rather than a pile of abandoned landing pages.
Accessible HTML, crawlable links, sensible page titles, descriptive headings, and page content that actually appears without requiring a search crawler to solve an escape room.

Google specifically recommends using the language people use to search for a topic in prominent page locations, including titles, primary headings, alt text, and link text. It also emphasizes crawlable links and applying established best practices for images, video, structured data, and JavaScript.

The implication is not “stuff exact-match keywords into every heading until the prose begins to sound like a ransom note.” The implication is that clarity has technical consequences. If your audience calls the service “corporate event video production,” do not hide that phrase beneath six paragraphs of brand mythology. If you serve Alpharetta, Atlanta, Buckhead, and the wider metro area, state that plainly where it is relevant. If a product integrates with Stripe, Shopify, React, or AWS, identify it precisely.

Machines cannot reward distinctions you refuse to make.

AEO Is SEO With Better Manners

AEO has become the preferred term because “SEO” carries baggage: affiliate content farms, keyword-stuffed blog posts, dubious link schemes, and the lingering smell of a conference room where someone has just said “dominate the SERPs.”

But the most useful interpretation of AEO is not as a replacement for SEO. It is SEO adapted to a world in which the user may ask a full question rather than type three nouns into a box.

People do not always search for “best video production company Atlanta.” They ask:

“How much should a two-minute brand video cost?”
“What should I ask a production company before hiring them?”
“What is the difference between a commercial, a brand film, and a customer testimonial?”
“Can a small business get usable social content from one filming day?”
“What kind of video helps a local service business generate leads?”

The answer-engine opportunity is to publish the useful explanation before a prospect asks an AI assistant to improvise one.

This does not mean every page needs to be written like a FAQ machine assembled by a nervous intern. It means that a strong page should make its main answer easy to locate. A reader should not have to excavate a conclusion from twelve scroll lengths of throat-clearing. Put the answer near the top, then earn the right to expand it with nuance, evidence, tradeoffs, examples, and practical detail.

The sweet spot is what might be called quotable completeness: a page with a concise, accurate answer that can stand alone, followed by enough depth that a skeptical human actually trusts it.

For example:

“For most small businesses, a one-day video shoot is most valuable when it produces a package—not one hero video. Plan for a primary brand or testimonial film, several short vertical clips, still frames, and cutdowns tailored for your website, email, paid social, and LinkedIn.”

That is a useful paragraph for a client. It is also a coherent candidate for retrieval and citation because it is specific, comprehensible, and bounded by a real-world context.

AEO rewards the same qualities a good editor rewards: directness, organization, source quality, and the confidence to say what is true without pretending it is true everywhere.

The Tangential Things: Helpful, But Not Holy

Then there are the tactics that can help, sometimes substantially, but should never become a substitute for having something worth finding.

Structured data belongs in this category. Schema markup can help search engines understand explicit facts about your organization, products, articles, videos, events, reviews, and local presence. It can also make you eligible for certain search appearance features when implemented correctly. But schema is not a magic wand. Adding FAQPage markup to a page filled with weak questions and vaporous answers does not transform it into expertise.

Use structured data to describe reality, not to decorate mediocrity.

The same goes for formatting. Headings, lists, comparison tables, author bylines, citations, definitions, summaries, and well-labeled images can improve readability and machine interpretation. They are good editorial practice. But they work best when they make an already-useful page easier to navigate.

Other tangentially useful tactics include:

Maintaining current business information across your website and major listings.
Publishing bylined content that establishes who is speaking and why they are qualified.
Showing your work rather than merely describing it: portfolio pages, before-and-after examples, case studies, process documentation, original photography, and video.
Making pricing logic, timelines, deliverables, exclusions, and policies intelligible.
Earning authentic mentions from clients, trade publications, local organizations, industry peers, and relevant communities.
Building topic clusters around genuine expertise rather than producing dozens of near-identical pages aimed at every conceivable variation of a phrase.
Improving page speed, mobile usability, accessibility, and navigation—not because milliseconds alone confer authority, but because a broken or frustrating page cannot become a trusted destination.

Google’s guidance explicitly warns against the fantasy that “AEO/GEO hacks” will replace sound SEO. It says site owners should prioritize useful, non-commodity content and effective SEO rather than tactics such as needless content chunking, inauthentic mentions, or special AI files created solely for generative-search visibility.

That phrase—non-commodity content—deserves to be framed and hung over every content calendar.

Commodity content is what happens when 1,200 websites publish an article titled “10 Tips for Choosing the Best Video Production Company.” The verbs change. The stock photography changes. The advice remains so bland that it could have been written by a corporate retreat.

Non-commodity content says something the world does not already have in identical form. It may be a breakdown of what clients misunderstand about video budgets. It may be a detailed postmortem on turning one conference shoot into 40 social assets. It may be a guide to filming in a particular city, with permit details, seasonal concerns, venue realities, lighting challenges, and vendor relationships. It may be a candid explanation of when a business should not hire a production company.

That last one is especially powerful. Real expertise occasionally tells the customer to save their money.

The llms.txt Situation

Now we come to llms.txt, the tiny text file that has inspired a large amount of digital hope.

The proposed idea is straightforward. An llms.txt file, placed at the root of a website, is written in Markdown and offers a concise description of the site plus a curated list of important pages and resources. In its ideal form, it acts as a clean map for language models: here is what this site is, here are the authoritative pages, here is the documentation, here are the policies, here are the most useful details.

This is not an absurd idea. A well-organized, machine-readable index is a reasonable courtesy to any system trying to understand a complicated site. Documentation-heavy products, developer platforms, software companies, large knowledge bases, and websites with sprawling resource libraries may find it useful as a form of information architecture.

But a courtesy is not the same thing as a ranking factor.

Google is unusually clear on this point: it does not use llms.txt or other special AI text files as a requirement for visibility in Google Search, AI Overviews, or its other generative search features. Google says that creating these files is unnecessary for appearing in its AI features.

That does not prove that no AI system will ever read llms.txt. It does mean that treating it as an indispensable AEO tactic is premature. There is no broadly validated evidence that adding the file reliably improves AI citations, referral traffic, answer inclusion, or search visibility.

So what works for llms.txt?

A restrained approach.

If you manage a technically sophisticated site, especially one with documentation, APIs, product guides, or a substantial library of evergreen resources, create one if it costs little to maintain. Use it as a curated index, not as a dumping ground and not as a mystical incantation.

A practical file should include:

A clear site or organization name.
A one- or two-sentence explanation of what the site does and who it serves.
Links to canonical, high-value pages.
Short descriptions explaining why each linked page matters.
Links to primary documentation, policies, product pages, editorial standards, or support resources where relevant.
Only pages you would be comfortable presenting as your best source of truth.

Do not use it to repeat every URL on your sitemap. Do not stuff it with promotional adjectives. Do not create it and then forget it while your site architecture, services, and pricing change around it like a shopping mall after an economic downturn.

And do not confuse it with robots.txt. A robots file communicates crawler-access preferences. An llms.txt file is a proposed content-discovery convention. They are different instruments, played in different keys.

For most small and medium-sized businesses, the better use of the next four hours is not creating an llms.txt file. It is improving the service page that has no clear deliverables, no evidence, no examples, no local relevance, no authorial authority, and no answer to the question a buyer is actually asking.

The Real Strategy: Become a Source

The anxiety around AEO comes from a legitimate concern: if AI answers the question, why would anyone visit the source?

The answer is that generic information becomes easier to summarize, while original value becomes more valuable.

An AI system can paraphrase “video builds trust.” It cannot convincingly replace your case study showing how a healthcare company used a patient-story series across recruitment, fundraising, and social campaigns. It can summarize generic wedding-video advice. It cannot replicate your location scouting, your production process, your visual taste, your local relationships, your original footage, or the judgment that comes from having stood in difficult light and solved the problem before the client noticed it existed.

The future-proof SEO strategy is not to write for a machine. It is to become the kind of source a machine—and more importantly, a discerning human—has reason to cite.

Build pages that carry a signature:

A clear author or organization behind the claim.
Evidence that the work comes from experience.
Details that could not have been invented from a dozen competing pages.
Transparent limitations and tradeoffs.
Current information, reviewed when it matters.
Language a customer understands without needing a decoder ring.
A structure that permits scanning without flattening the subject into mush.

The great irony of answer engines is that they may push the web toward a simpler editorial truth. The content that survives will not be the most “optimized” in the vulgar sense. It will be the content that gives the best answer, proves it knows what it is talking about, and leaves behind the unmistakable impression that an actual person—or a serious organization—was there.

That is not a loophole. It is the job.