LLMs Really Value Unique Content

Be Human, Make Mistakes

That is the short answer SCAILE CEO Simon Wilhelm gave host Casey Cheshire on the Connect to Market podcast when asked the single most important thing marketers should do to show up in AI search. The reasoning is simple: LLMs mostly reproduce what is already out there, so the only durable way to be cited is to publish material the models have not seen before, written by a human who has earned the right to say it.

The instinct in most marketing organisations is the opposite. Polish every sentence. Strip out the personality. Match the brand voice template. That sterilising process makes content easy to QA and impossible to cite. AI assistants reward distinct angles, original phrasing, and the kind of opinions only a real practitioner would commit to writing.

In the hour-long episode, Simon walks through the practical implications of that for marketing leaders. This article expands on the main themes and translates them into tactics you can apply this quarter.

Why LLMs Reward Human Voice

Large Language Models have ingested most of the public internet. They have seen the templated thought-leadership posts, the same SaaS landing-page copy repeated ten thousand times, and the AI-generated content from every competitor that bought a writing tool in the last two years. When a buyer asks a question, the model is looking for content that adds new signal to what it already has.

A page that says something the model has not seen before, written in a voice that does not pattern-match to a content factory, gets weighted more heavily. It does not need to be longer. It does not need to win at backlinks. It needs to contribute a non-obvious point of view that the model can quote.

That is why "be human, make mistakes" is not a soft sentiment. It is a strategy. Mistakes, in this context, means writing what you actually think rather than what the template wants. Phrasing things in ways that are recognisably yours. Including the experience that only your team has. Showing up as a person, not as a content calendar.

The companies that do this well end up with a moat. Other brands can copy the template, but they cannot copy the lived experience.

SEO, AEO, GEO - The One-Minute Definition

In the conversation, Simon clears up the alphabet soup the industry has bolted onto AI search:

• SEO is the classic discipline: getting a page to rank in traditional search. Still the foundation; still where most discovery starts.
• AEO (Answer Engine Optimization) is about becoming the snippet the LLM quotes when it synthesises an answer for the user. AEO is tactical: it lives in headings, structured data, FAQ blocks, and crisp comparison tables.
• GEO (Generative Engine Optimization) is broader: a company-wide competency that lines up marketing, sales, customer success, and product around a coherent message so AI search represents your brand the same way everywhere. GEO is strategic.

Good SEO is still the foundation. AEO and GEO sit on top of it. Think of it as three concentric circles: SEO gets you into the index, AEO gets you into the answer, GEO makes sure the answer matches what you want the market to hear.

The biggest mistake operators make is treating these as alternatives. They are layered. A company that does AEO tactics without an SEO foundation gets sporadic citations that do not compound. A company that does GEO strategy without AEO execution has a beautiful narrative that AI assistants never repeat. All three need to work together.

The Timeline Is Faster Than Anyone Expected

The biggest surprise for marketers used to waiting 3 to 6 months for SEO results: with AEO, results show up in 6 to 14 days, sometimes the next day.

Simon describes how SCAILE measures it: tracking prompts across ChatGPT, Claude, Perplexity, and Gemini before and after publishing, and what that compressed feedback loop changes about how a content team operates.

A few practical implications:

• Plan in two-week cycles, not quarters. AEO investments validate fast enough that a quarterly content calendar is too slow. Two-week sprints, with explicit prompts to win and explicit measurement at the end, are a better fit.
• Ship before the page is "perfect". With a 14-day feedback loop, the cost of waiting two extra weeks to polish a page is the loss of an entire iteration cycle. Publish at 80 percent quality, measure, iterate.
• Kill underperformers fast. A page that does not earn AI citations in three weeks usually never will. Either rewrite it or take it down. Holding onto under-performers in the hope they will compound is a habit from the SEO era.
• Reward the team for citations, not posts. The metric that matters is "how many AI assistants now name us for the prompts we care about". Output counts (number of articles shipped) lag that metric.

This speed of measurement is one of the genuinely new things AI search introduces. Marketers used to operating in long, opaque cycles need to retrain themselves to a tighter loop. It is uncomfortable, and it is also one of the largest sources of unfair advantage available right now.

Tactics from the Conversation

The episode is dense with specific tactics. The four that stood out:

• Refresh before you publish: update facts, add tables, FAQs and internal links to existing pages. Typical lift: ~40%, zero new content. Most marketing teams have years of buried content that is two updates away from being citation-ready. Refresh first; write new only when the existing inventory has been exhausted.
• Use Bing Webmaster Tools: Microsoft now shows which AI search queries you appear in. Free, and Google does not provide the equivalent. Microsoft is the upstream of ChatGPT's search citations, so this is one of the highest-signal free dashboards available right now.
• Watch AI crawlers in Google Analytics: live since 13 May 2026 - see exactly which models read your site. GPTBot, ClaudeBot, PerplexityBot, and others all hit your site differently. Knowing which models actually crawl you is the difference between guessing and optimising.
• Optimise per LLM: different audiences live in different models. B2B SaaS for engineers? Optimise for Claude. Mass-market consumer brands? ChatGPT first. Research-heavy verticals? Perplexity. Treating every LLM as the same audience is the same mistake as treating every social platform the same.

The 30/60/90 Plan for AI Visibility

If you are starting from zero, the conversation suggests a rough sequence:

Day 1 to 30: Baseline and quick wins.

• Run an AI visibility audit (free tool at scaile.tech/tools/ai-visibility).
• Identify the top ten buyer prompts your category should own.
• Audit your existing content against those prompts; mark the pages that are within striking distance.
• Refresh the top five within the first two weeks: tables, FAQs, sharper opening, internal links.
• Measure week-over-week. Document what changed.

Day 31 to 60: Production at velocity.

• Build out a content production loop that ships at least one citation-targeted page per week per prompt.
• Set up tracking for the top twenty prompts. Move the needle on at least half.
• Start instrumenting AI crawler traffic in Google Analytics 4.
• Begin per-LLM optimisation: identify which model your highest-intent traffic comes from and write specifically for it.

Day 61 to 90: Compound and systemise.

• Move from "writing pages" to "owning a topic cluster". Each cluster should win a coherent set of prompts.
• Convert the measurement loop into a weekly business review with sales attached. Citations should connect to pipeline.
• Decide where to double down (which prompts to defend, which categories to expand into) and where to step back.

Most teams underestimate the speed at which results appear, and overestimate the volume of work required. The right number of pages is usually smaller than expected; the right cadence is faster.

Why Brand Voice Matters More Than Ever

A theme that runs through the episode: AI search amplifies the consequences of having a distinctive voice, and amplifies the cost of not having one.

A brand without a clear voice gets paraphrased into the median description for its category. That description is bland, generic, and indistinguishable from competitors. The buyer reads the AI's answer, sees four vendors described in similar terms, and picks the cheapest or the most familiar.

A brand with a clear voice gets quoted verbatim. The buyer reads phrases that signal a real point of view, recognises them as different, and pays attention.

The implication: brand investment used to feel like a long-term, hard-to-measure cost. In AI search it is suddenly a fast, measurable performance driver. The companies that have spent the last few years sharpening how they talk about themselves are the ones AI assistants quote.

The Bing Webmaster Tools Tactic, Explained

Of all the tactics Simon mentions, the Bing Webmaster Tools tip is the most under-used. Most marketers either ignore Bing entirely or check it once a year. That is a mistake.

Bing is upstream of ChatGPT's search citations. When ChatGPT does a web lookup for a query, it often draws from Bing's index. That means the queries your site appears for in Bing Webmaster Tools are a leading indicator of the AI search queries you will appear for.

The dashboard is free. The data is detailed. Set it up, set a fortnightly review, and treat it as your earliest signal of AI search momentum.

How to Pick Which LLM to Optimise For

Optimise for the model your buyers actually use. A quick decision framework:

• Mass-market consumer: ChatGPT first. Massive reach, conversational default, lowest friction for non-technical users.
• B2B and technical: Claude first. Strong in research, coding, and longer-form analysis. Used disproportionately by engineering and product teams.
• Research-heavy and citation-driven: Perplexity first. Surfaces sources prominently; strong for analysts and consultants.
• Anywhere Google still owns discovery: Google AI Overviews. Hard to ignore even as ChatGPT grows.

The right answer for most companies is "all four, with different weights". The weights depend on where your buyers actually live, not on aggregate market share charts.

Frequently Asked Questions

Do I need to rewrite all my content for AI search?

No. Most teams should refresh before they rewrite. Updating facts, adding a FAQ block, including a comparison table, and tightening the opening paragraph generates a ~40% lift on average. Only write new content when the existing inventory has been exhausted for the prompts you care about.

How fast does AEO actually work?

Six to fourteen days for most pages and prompts. Some show within 24 hours; others take a full sprint. The variance comes from prompt competitiveness and how well-structured the source page is. Plan in two-week cycles.

Is Bing Webmaster Tools really useful in 2026?

Yes. Bing is upstream of many ChatGPT citations. The queries you appear for in Bing are a leading indicator of the AI search queries you will appear for. It is free and the data is detailed. Set up a fortnightly review.

How do I track AI crawler traffic?

Google Analytics 4 has shown AI crawler traffic since 13 May 2026. You can also use server-side logs. Watch for GPTBot, ClaudeBot, PerplexityBot, CCBot (Common Crawl), and Bingbot's AI variants. Knowing which models actually crawl your site changes which model to optimise for first.

Can AI-generated content rank in AI search?

Sometimes, but it is a fragile strategy. LLMs increasingly detect their own output and weight it lower. Hybrid workflows where a human writes the angle and AI helps structure and copy-edit perform better than pure AI generation. Pure templated AI content is the easiest thing for the next model update to filter out.

What metric should our content team report on?

Citation rate (how often your brand appears in AI answers for the prompts you care about), share of voice (your brand mentions divided by total brand mentions across those prompts), and accuracy (how often the AI's description of you matches your positioning). Track weekly, review with sales attached.

Watch the Full Episode

youtube.com/watch?v=DqonQJutUgI