The SEO Tactics That Still Work in AI Search (And the Ones That Are Dead)

A founder messaged me last week with a screenshot of his SEO agency's retainer. Line item after line item for tactics that stopped working three years ago: monthly "blog comment outreach," a Domain Authority target, keyword density reports. He was paying good money to chase ghosts while his competitors were getting cited in AI Overviews and ChatGPT answers.

The rules changed when answer engines started summarizing the web instead of just linking to it. Some tactics survived that shift and got stronger. Others quietly died and nobody told the agencies still billing for them. Here is the honest split, so you can stop funding the dead ones.

Still works: original data and research

If you publish a number no one else has, everyone who writes about that topic has to point back to you. That was true in 2015 and it is more true now, because AI answer engines are hungry for specific, citable facts. A model summarizing "average cost of X" wants a source with a real figure, and it favors the page that produced the data over the ten pages that quoted it.

You do not need a research department. A survey of 200 customers, a teardown of 50 competitor pages, a before-and-after from your own account. One genuine data point, published cleanly, earns links and citations for years.

Still works: digital PR

Getting mentioned on sites people actually trust is the closest thing to a rankings cheat code that is still legal. A link from a real publication carries authority, sends referral traffic, and teaches answer engines that credible sources associate your brand with a topic.

The bar is higher than a mass pitch. Tie your outreach to the original data above, give a journalist a reason to care, and one placement can outperform a hundred low-quality links.

Still works: quality backlinks and citations

Here is where people get confused. "Link building is dead" gets repeated every year, and every year links keep predicting who ranks. What actually died is the junk version: spun directories, private blog networks, comment spam.

Real links from relevant, trusted sites still carry weight, and getting listed on the right directories and launch platforms is one of the cleanest ways to earn them. Those listings do double duty in the AI-search era. They are backlinks, and they are structured citations that tell answer engines who you are and what category you belong to. That is exactly the job BacklinkBot does: it submits your product to 100+ relevant directories and listing sites so you build real links and consistent citations without the manual grind.

Still works: topical authority

Owning one topic beats ranking for one keyword. When your site covers a subject deeply, from the beginner question to the expert edge case, both Google and AI models start treating you as the reference for that space.

Think in clusters, not keywords. Pick a topic you can genuinely own, then build the full map of pages around it and link them together. Depth is the signal now.

Still works: answering in the first 100 words

Answer engines lift the passage that responds to the question fastest and cleanest. If your page buries the answer under 800 words of throat-clearing, a competitor who leads with a crisp two-sentence answer gets pulled into the summary instead.

Lead with the answer, then earn the reader's attention for the depth below it. This single habit is one of the highest-leverage changes you can make for AI visibility right now.

Still works: reviews with real substance

Reviews still count, and descriptive reviews count more. A wall of five-star ratings with no words tells an answer engine nothing. Reviews that mention specific features, use cases, and outcomes give models the language they use to match you to a searcher's intent.

Ask happy customers to say what they used you for and what changed. That specificity is the part that ranks.

Dead weight: stop paying for these

• Blog comment "links." Nofollowed, ignored, and a spam signal. This tactic died a decade ago.
• Keyword stuffing. Repeating a phrase to hit a density target reads like a robot to both people and models. Write for the human.
• Chasing a single keyword. Ranking for one term while ignoring the topic around it leaves you fragile and invisible in AI answers that span the whole subject.
• Domain Authority as a goal. DA is a third-party score invented by a tool company, not a metric Google uses. It is a rough external estimate, useful for a quick gut check and nothing more. Optimizing your DA number is optimizing someone else's marketing metric.
• Page speed as an obsession. Speed matters when your site is genuinely slow and losing users. Once you are in a reasonable range, shaving another 100 milliseconds will not out-rank a competitor with better content and stronger links.

Where to spend instead

Move the budget you were burning on dead tactics into the four that compound: publish original data, earn real links and citations, build topical depth, and answer questions fast. Those are the inputs answer engines reward, and unlike a DA score, they are things you actually own.

If earning quality links and citations is the piece you keep putting off, that is the exact grind BacklinkBot automates. Submit your site to 100+ handpicked directories once, and start building the links and citations that AI search actually reads. Start your submissions here.