Every 'Innovative' SEO Tactic We've Seen Was a Repackaged Basic

Eight tactics that are old work in a new jacket

We've been pitched, sold and occasionally seduced by every one of these. Each is real work worth doing. None of them is new, and the renaming is doing a lot of commercial labour that the underlying idea never asked for.

The test is simple: if you can describe a tactic to a marketer from 2015 and they nod rather than blink, it isn't new.

Why the renaming keeps happening

It isn't a conspiracy. It's an incentive structure, and once you can see it you can price around it.

A named new tactic is a reason to send an email, book a call, justify a rate rise, and fill a conference slot. 'We're going to keep doing the same competent work for another eleven months' is a true sentence that has never once won a pitch. Nobody has ever been given a stage for the eighteenth consecutive month of a publishing cadence, which is a shame, because that's where the compounding lives.

• Agencies with churn need novelty. If clients leave at month nine, you need something that makes month nine feel like month one. A new tactic is the cheapest way to manufacture that feeling.
• Tool vendors need categories. Naming a practice creates a product line, a comparison page and a search term nobody else owns yet. That's a rational business decision and it's why half the vocabulary exists.
• Conferences need programming. Forty slots a year, and 'the basics, again, correctly' fills none of them.
• Founders want it to be true. A new tactic implies your flat quarter wasn't your fault — the rules changed. That's a far more comfortable story than 'we published four thin pages and stopped'.
• And yes, we're implicated. We're a marketing agency writing marketing content. The temptation to name something and claim it is constant. This article is us resisting it on a good day.

The first genuinely new thing: entity consolidation

Here's what changed, and why it's different in kind rather than degree.

Classic SEO optimised a URL against a query. Everything you touched, you owned. AI assistants and AI Overviews don't work that way — they assemble a description of your company from everything written about it anywhere, then answer from that composite. The unit of work moves from the page to the entity, and most of the entity lives on properties you don't control.

The honest ancestor here is local SEO's obsession with consistent name, address and phone across directories. We're not going to pretend that's unrelated — it's the same instinct. What's different is the scope and the reader. It now applies to every factual claim about you: what you sell, what you charge, who you serve, which cities you operate in, who runs the company. And the reader is a language model, which doesn't shrug at contradictory sources. It picks one, states it confidently, and repeats it to your prospect.

The second: query fan-out, and what it does to coverage

Google has described its AI Mode as using a query fan-out technique — taking one question, issuing several related searches across subtopics and sources, then synthesising an answer from the results. Assistants do their own version of the same thing.

That changes what 'ranking for a query' means. A single user question becomes several machine queries you never see. Your page can be the best answer to the question as asked and still be absent from the response, because the citation went to whoever answered sub-question three about pricing, or the exception, or the comparison against the obvious alternative.

So coverage stops meaning keyword variants and starts meaning decomposition. Take the question your page targets, break it into the six or eight things a careful person would want to know underneath it — how much, how long, what breaks, compared to what, what if I'm in this specific situation — and answer each one, in its own section, self-containedly. Passages start to matter more than pages, because a passage is what gets retrieved.

The uncomfortable consequence is measurement. You cannot see the fan-out queries in Search Console. You often can't see the original question either. The clean feedback loop that made SEO the most accountable channel in marketing is partially broken for this surface, nobody has a good substitute yet, and anybody selling you precise attribution for AI citations is selling you a modelled number with a confident face on it. We'd rather say that than invoice for it — the broader argument is in is generative engine optimization real.

The things we're not doing yet, and why

Restraint is a tactic too, and it's the only one that gets cheaper the longer you hold it.

• [llms.txt](/glossary/llms-txt) — a proposed convention for telling AI systems what's on your site. We've seen no confirmation that the major assistants actually consume it. It takes about an hour to add, so add it if the hour is free. Don't build a plan around it and don't pay anyone a retainer line for it.
• Publishing at volume to win citations. Volume without differentiation is exactly the pattern Google names as scaled content abuse, and the penalty lands on the whole domain rather than the thin pages. We drew the line in programmatic SEO and the scaled content line.
• Schema on everything. Structured data helps machines parse what's already there. It does not manufacture authority, and no amount of markup makes a weak page a strong one.
• A separate 'AI visibility' retainer. Ask what's in it. If the answer is clear factual pages, consistent entity data, structured answers and earned mentions, you're being quoted twice for work that was already in the SEO scope.
• Chasing every core update with a rewrite. Rankings wobble for a fortnight after an update while the index settles. Rewriting in week one means rewriting against noise.

How to test a new tactic without risking the site

When something genuinely new does turn up — and twice in three years, something has — you want to try it without betting the domain on a stranger's blog post. Six rules, and we hold to them on our own site first.

1. Ring-fence it. One directory, or 5–10% of URLs. Never your top twenty pages by revenue, never the templates those pages share. If the tactic is site-wide by nature, that's a reason for more caution, not less.
2. One variable, one window. Change the template and the publishing cadence in the same month and you've learned precisely nothing at the end of it.
3. Write the kill criteria before you start. Metric, threshold, date. 'Let's see how it goes' is how a test quietly becomes a permanent feature nobody can justify.
4. Keep the rollback. Version the old template, keep the old copy, and use noindex before you delete so a reversal doesn't cost you the URL and its history.
5. Measure against a control. Comparable untouched pages in the same section. Site-wide traffic in a month that contained a core update tells you about the update, not about your test.
6. Cap it in time. Ninety days, then a decision: keep, kill, or extend once with a stated reason. Tests that run indefinitely aren't tests, they're habits.

Three questions we ask before billing anyone for something new

Cheap filters, applied in about ninety seconds, on a call.

1. What's the old name for this? If there is one, price it as the old thing and check whether your previous agency already did it. If there genuinely isn't one, keep listening — that's rare and interesting.
2. What breaks if it's wrong? Reversible on twenty pages is a test. Irreversible across the site — a URL restructure, a mass deletion, a migration — is a bet, and bets get made with evidence rather than enthusiasm.
3. Who profits from me believing this? A tool vendor defining a category, an agency with a churn problem, a conference with slots to fill. None of those make the claim false. They do tell you how much scepticism to bring.

What we'd spend the hour on instead

If you have one hour and a choice between reading about a tactic named this quarter and doing something dull, do the dull thing. In order of what usually pays fastest:

Pull your striking-distance list out of Search Console and fix the top five pages. Add internal links from your strongest pages to the ones you actually want ranked. Delete or merge everything that hasn't earned a visit in a year. Test your site on a mid-range Android phone on a 4G connection, because that is what most of India is browsing on, and fix whatever takes more than three seconds. Ring one customer, record the call with permission, and use their actual words instead of the ones in your brief.

None of that will get you invited to speak anywhere. All of it compounds, and the compounding is the part competitors can't copy by reading a slide. Whether covering a subject properly is even a real ranking property is worth arguing about too — we did, in does topical authority actually exist.

We take three clients a month and we guarantee movement against your own frozen trailing-90-day organic lead baseline, not a position on a page. That constraint is why we're allergic to novelty: an unproven tactic costs us free months, not you. It turns out the fastest way to make an agency conservative about tactics is to make it liable for the results.