# SEO advice is often unproven. Check the evidence first - Episode: P1E02 (Part 1: Modern SEO) - YouTube: https://www.youtube.com/watch?v=WKoE1X25kqw - Page: https://irongoo.com/videos/seo-evidence - Duration: 07:26 - Published: 2026-10-23T14:30:00+05:00 Learn how to judge an SEO claim before you spend time or money on it. See which popular tactics have evidence behind them and which remain unproven. ## What this episode covers - Seven levels of evidence, from Google's internal API documentation to SEO folklore. - Five claims about LSI keywords, schema, author bios, word count and AI content. - Why rich results and extra clicks do not necessarily mean higher rankings. - What controlled tests can tell you about changes in organic traffic. - How to use the printable evidence ladder and claim checker worksheet. ## Key moments - [00:00 Intro](https://irongoo.com/videos/seo-evidence?t=0) - [00:51 The evidence ladder](https://irongoo.com/videos/seo-evidence?t=51) - [02:43 Claim 1: LSI keywords](https://irongoo.com/videos/seo-evidence?t=163) - [03:21 Claim 2: Schema boosts rankings](https://irongoo.com/videos/seo-evidence?t=201) - [04:10 Claim 3: Author bios](https://irongoo.com/videos/seo-evidence?t=250) - [04:58 Claim 4: Word count](https://irongoo.com/videos/seo-evidence?t=298) - [05:27 Claim 5: Google hates AI content](https://irongoo.com/videos/seo-evidence?t=327) - [05:59 Reading a case study](https://irongoo.com/videos/seo-evidence?t=359) - [07:00 Wrap up](https://irongoo.com/videos/seo-evidence?t=420) ## Transcript ### Intro Here is an uncomfortable fact about SEO. When one of the largest SEO testing platforms ran controlled experiments on real websites, more than half of the changes they tested made no detectable difference to traffic. Not better. Not worse. Nothing. Think about what that means. Most of the tactics you read about in blog posts, threads and videos have never been tested at all. And when someone does test them properly, the honest result is often a flat line. So in this episode, we are not going to give you another list of tips. We are going to give you something more useful: a way to judge any SEO claim before you spend money on it. We call it the evidence ladder. Then we will put five popular claims on it, live, and give each one a verdict. ### The evidence ladder Here is the ladder. Seven rungs, strongest at the top. Every claim you hear about SEO sits on one of them, and where it sits tells you how much weight to give it. At the top: things Google stores. In 2024, thousands of pages of Google's internal API documentation leaked. They show which signals Google keeps, for example click data and a site level authority score. They do not show how much each one weighs, but they are the closest thing we have to looking inside. Second: the court record. In the United States antitrust case against Google, engineers testified under oath and internal documents were filed as evidence. Courts check that record, which is why it sits so high. Third: things that were measured. Controlled split tests, where one group of pages changes and a matching group does not, and large studies across thousands of sites. This is where you find out whether a tactic actually moves traffic. Fourth: what Google says. Google's documentation is very reliable about what Google wants and what breaks its rules. It is less reliable as a description of how ranking works inside, because that is not what it is written for. Fifth: field work. What experienced practitioners see work again and again across real campaigns, including ours. When a point comes from our own campaigns, you will see it badged as our field work. Sixth: patents. Google files patents for ideas it may or may not use. A patent tells you what Google invented. It does not tell you what is running today. And at the bottom: folklore. Claims that get repeated so often they feel true, with nothing underneath them. One rule before we start. A claim that lives on the bottom two rungs is not automatically false. It is just unproven. Treat it like a hypothesis, and never bet your whole budget on a hypothesis. ### Claim 1: LSI keywords Claim number one. You need L.S.I. keywords. The idea is that Google uses a technique called latent semantic indexing, so you should sprinkle related terms into your pages. Latent semantic indexing is a real technique, from the late nineteen eighties. It was designed for small, fixed collections of documents, not the web. In 2019, Google's own search advocate said it plainly: there is no such thing as L.S.I. keywords. No test, no leak and no testimony supports the idea. Writing naturally about a topic does use related words, but that is good writing, not a secret Google technique. ### Claim 2: Schema boosts rankings Claim number two. Adding schema markup boosts your rankings. Schema is the structured data you add to a page's code to describe what is on it. Here the evidence is interesting. In controlled tests, schema has helped traffic, but through rich results, the stars, prices and extra lines that change how a result looks. That changes how many people click. It is not the same as ranking higher. And when Google removes a rich result, the win goes with it. FAQ rich results were removed for every website in May 2026. Google's own 2026 guide also says there is no special markup you need to appear in its AI answers. So the verdict is split. Schema works for rich results your page can actually win. As a ranking boost, it is a placebo. ### Claim 3: Author bios Claim number three. Add author bios and bylines, and your rankings will rise, because Google rewards expertise. This one was tested directly. A site added a byline with the author's name and photo at the top of its pages, and a full bio at the bottom. The result: no detectable impact on organic traffic. They changed the design and tested again. Same result. And in the 2024 leak, there is no author reputation score at all. There is only a flag that marks which entity wrote a document. Does that mean authors do not matter? No. Real expertise shows up in the content itself: first hand experience, accurate facts, things only someone who did the work would know. A bio box cannot fake that. It is a label, not the thing. ### Claim 4: Word count Claim number four. Longer content ranks better, so every article should be at least two thousand words. Google answers this one in its own documentation, as a question and answer. Does Google have a preferred word count? No, we don't. The correlation people point to is real but misleading. Thorough pages tend to be long, and thorough pages tend to rank. The length is a side effect, not the cause. Padding a page to hit a number usually makes it worse. ### Claim 5: Google hates AI content Claim number five. Google penalises AI content, so anything written with AI will be buried. Google's spam policy is about pages made mainly to manipulate rankings at scale, no matter how they are created. Its guidance says using generative AI to produce large amounts of text without oversight is little to no effort. Read that carefully. The problem it names is scale without oversight and value, not the tool. Thin pages written by people at scale fail for the same reason. ### Reading a case study Now, the thing you will see most often in SEO is not a claim. It is a case study. So here is one of ours, straight from Search Console. An electronics spare parts store. On the ninth of August 2026, before we started working with it, Google sent the store forty clicks. On the sixteenth of October, it sent two hundred and four. Five times the traffic from search. And visibility grew with it. On those same two days, the store appeared in Google results one thousand and sixty six times, and then three thousand and twenty six times. Nearly three times as many people seeing it, and five times as many clicking through. It holds across whole months too. Average daily clicks went from seventy nine in August, the month we started, to a hundred and twenty in September. That is the kind of evidence that sits high on the ladder: measured, dated, and straight from the source. And it is how we will show results in this whole series. ### Wrap up Here is the whole episode on one screen. Five claims, five verdicts. Most of what gets repeated in SEO sits on the bottom of the ladder. The things that do work tend to be boring: clear pages, real expertise, and changes you can measure. The evidence ladder is free as a printable card, along with a claim checker worksheet, on the episode page at irongoo.com. Next time someone sells you a tactic, put it on the ladder first.