Czy SEO jeszcze ma sens? Senuto o spadkach ruchu, AI Mode i widoczności w ChatGPT | Artur Jabłoński

Main topic

The episode discusses whether SEO still “makes sense” in Poland amid Google’s AI search features (especially AI Mode) and broader LLM-driven visibility. It uses Senuto’s research (and follow-up analysis) to quantify traffic losses, how much can be recovered, and what “the new SEO” should optimize: citations/mentions in AI answers, not just classic rankings and clicks.

Key findings: traffic loss and recovery

• After Google’s AI-related changes, organic traffic declined by ~20% in Senuto’s study.
  • The decline is described as worsening/accelerated in stages: I/O change first, then AI Mode “completed” the damage.
• The decline is not stable:
  • it changes month-to-month with technological changes and differs across categories.
• Recovery is possible but limited:
  • Senuto claims they can recover about 60% of the lost traffic (i.e., recover part of the lost ~20%).
  • The host emphasizes this does not mean full compensation, because it’s 60% of a reduced baseline, not a return to prior levels.
• The biggest concern: profitability and CTR
  • The tool/product owner argues pure SEO has become less profitable due to higher entry cost (time/money) and lower returns and CTR.
  • Blogging/own-site content is viewed as increasingly hard to justify purely for traffic acquisition (at least in the studied contexts).

What matters in AI search: citations and their order (not just top rankings)

The discussion shifts from “traffic from Google” to how AI answers cite sources:

• Senuto’s analysis comparing top 10 results vs. AI answer citations suggests:
  • AI Mode: ~34.7% of top-10 domains are quoted.
  • Google I/O (IO equivalent in transcript): ~42% are quoted.
• Citations extend beyond the top 10:
  • Even when top performers are often quoted, a substantial portion of quoted sources comes from outside the top 10.
• The episode argues that position within citations matters, not just being quoted:
  • If you are cited first/earlier, you’re likely to earn faster clicks and more CTR.
• Therefore, “the new SEO” becomes:
  • compete for citation placement in AI responses, not only SERP position.
  • treat AI answers like a funnel where users may decide before clicking a traditional result.

Quality of traffic from LLMs vs. Google (and measurement limits)

• A point raised: Google reps claim less traffic can be higher quality.
• Senuto’s data cannot fully validate this:
  • they don’t have direct “traffic quality” evidence comparable to Google Analytics + Search Console.
  • they suggest it logically could be higher because AI responses do more of the decision-making internally, potentially reducing “unfinished” clicks.
• They observe that qualified queries may be higher for LLM-originated sources (anecdotal/client confirmations), but stress it’s not comprehensive research.

Platform differences: no single “one-size-fits-all” strategy

A key conclusion: each LLM/assistant behaves like its own ecosystem, with different citation habits and source preferences.

Examples from the episode:

• GPT tends to reuse trusted/user-generated sources (e.g., Reddit/YouTube-like sources are emphasized).
• Perplexity is more e-commerce/blog oriented.
• Copilot lacks citations, which changes the influence dynamics.
• Gemini is described as more conservative and harder to influence because it relies more on training data.

Branding/visibility measurement should be platform-specific, ideally with weighted evaluation—otherwise a “100% visibility” metric on one platform may be misleading if other platforms show much less.

Uncertainty and non-determinism: measurement is hard

• LLM outputs are non-deterministic:
  • brand order changes frequently,
  • the same public query can return different “models” or variants.
• Senuto reports:
  • only ~30% of brands stay consistently in first place over time;
  • outside top-3 brands, results behave more like a lottery.
• Measurement should focus on:
  • presence frequency (share of responses containing your brand),
  • “model first choice” / recommendation behavior,
  • rather than only absolute rank.

When to invest: not the end of Google, but selective readiness

The episode challenges the “apocalyptic” narrative that LLM chat will immediately replace search:

• Chat/LLM traffic (described as ~4%) is said to be not growing fast enough to justify broad investment by small businesses.
• Larger brands may need to act earlier due to scale, monitoring costs, and internal pressure.
• The key question becomes whether current LLM visibility work is hype vs. real business need, given how quickly platforms change.

Ethical visibility vs. spammy tactics (and “ranking” hacks)

A late segment addresses unethical practices:

• The guests discuss fake/manipulative “rankings” where agencies place themselves first to generate citations in LLM answers.
• Senuto claims many such tactics are prevalent (they mention ~90% of these rankings in a studied space look like this), and they expect a legal/purge cycle over time.
• They caution that some tactics may have short shelf life, but acknowledge they currently work.

“Ethical” recommendations for increasing citations

Instead of manipulation, the episode suggests practical content intentions:

• Use six “intent” types (described generally), especially:
  • definitions, comparisons, use cases, and diversified angles.
• Write as if the model doesn’t already know your brand:
  • build mentions in a brand-contextual way (entity + relation),
  • diversify formats to improve frequency and breadth of mentions/citations.
• Technical/format guidance:
  • improve faster model processing (accessibility, structured data placed in the head section),
  • structured data helps models extract information without heavy “scraping,” increasing citation likelihood.

Senuto product direction (monitoring tool)

• Senuto says it’s building an AI visibility monitoring tool (“monitor”):
  • focuses on how the model sees a brand,
  • measures citation gaps,
  • identifies opportunities where publishing could increase the chance of being cited.
• The tool is initially for enterprise clients, with consumer availability planned.

Presenters / contributors

• Artur Jabłoński (host)
• Maciej Chmurkowski (Senuto product owner; main guest/expert)
• Damian Sołkowski (mentioned as conducting/author of study based on Senuto data)
• Senuto / Senuto team (implicitly referenced regarding ongoing research and product)