Retrospectivă: Search Evolution Summit (2026.09.03-04, Cluj-Napoca) - prima conferință SEO din Transilvania

15 Sep, 2026

Pe 3 și 4 septembrie 2026 a avut loc la Cluj-Napoca a doua ediție a celei dintâi conferințe SEO din Transilvania - Search Evolution Summit. Mai jos, notițele mele, alături de imagini de la eveniment. Am făcut și scurte clipuri video cu telefonul.

Cuprins

Clipuri video din cadrul conferinței, pentru atmosferă
Fotografii
Notițe

(top)

Clipuri video din cadrul conferinței, pentru atmosferă:

Vezi video pe YouTube Shorts »

@olivianbredaSearch Evolution Summit (2026.09.03-04, Cluj-Napoca) - prima conferință SEO din Transilvania ♬ original sound Olivian Breda

Vezi video pe TikTok »

View this post on Instagram

A post shared by Olivian Breda (@olivianbreda)

Vezi video pe Instagram » (top)

Fotografii:

Album foto pe Facebook – partea 1 » Album foto pe Facebook – partea a 2a » Album foto pe Facebook – partea a 3a » Album foto pe Google Photos » Postare Instagram #1 » Postare Instagram #2 » (surse foto - camera foto proprie, telefon propriu, fotografii furnizate de organizatori) (top)

Notițe

(top)

Notițele mele de la eveniment - Ziua #1

(top)

Multimodal Video Optimization - Myriam Jessier (Myriam Jessier: AI search strategist)

While a text sentence contains very little data, a 5-minute video at 60 fps provides a massive volume of verifiable evidence across visuals, nuanced audio, and a full text transcript.

Text provides minimal data: Isolated sentences lack sufficient information density.
Video delivers multi-layered evidence: High frame counts paired with audio and transcript streams generate substantial, verifiable detail.

Text offers almost no verifiable footprint for authenticity, while rich media like video provides thousands of cross-referencing signals—visuals, audio, and transcripts—that make human origin and accuracy far easier for AI to verify.

Text lacks unique proof: Short copy gives algorithms almost no context to distinguish human creation from AI generation.
Video provides a massive data footprint: Thousands of visual frames combined with audio dynamics create a deep, layered "surface area."
Multimodal cross-referencing enables trust: The more data streams available to cross-check, the easier it is for systems to validate authenticity.

When an AI system processes a video, it breaks the file down into 3 streams: visual, audio, and text. Three layers in content recognition:

The visual layer (object detection and OCR - Optical Character Recognition) to identify physical products, faces, and read any text;
Automatic Speech Recognition (ASR) to transcribe spoken words into text, while also analyzing the speaker‘s vocal tone, cadence, and sentiment;azazsss
The text alyer - the AI relies on structured metadata and human-verified transcripts as a bridge to read and understand the video‘s semantic meaning.

Because Gemini 1.5 Pro defaults to sampling video at 1 frame per second, brief visual elements—such as a product appearing for only half a second—are likely to be missed during tokenization. How to avoid AI bots analyzing not each frame, but looking in several steps at videos? Keep critical visual information, like text overlays or product shots, static on screen for at least 3 seconds. Think of the 3-second rule as your insurance policy against product blur. To avoid errors in AI speech detection, you should enunciate:

Clearly enunciate key brand entities and pause slightly before and after saying them.
Vary your vocal emphasis like a highlighter, so the AI knows which points matter most.

TikTok indexes spoken audio in real time, so speaking your main keywords clearly in the first three seconds and reinforcing them with contextual terms is critical for proper product categorization.

Front-load primary keywords: State key terms clearly within the first 3 seconds so the real-time transcription captures them immediately.
Establish context: Use related vocabulary throughout the video to build semantic relevance.
Improve classification: Giving the algorithm clear, spoken context ensures accurate product indexing and targeting.

You should be redundant: Say the product name repeatedly, not "it" or "this," so transcription and entity ranking lock into it. Direct subject-predicate-object sentences form clear semantic triples that LLMs can extract into knowledge graphs, whereas vague marketing prose dilutes vector meaning and prevents citations.

Use explicit syntax: Subject-predicate-object structures provide the precise triples LLMs need to reliably map facts.
Avoid flowery marketing copy: Abstract or vague phrasing weakens vector embeddings.
Maximize citation potential: Clear, unambiguous factual statements ensure models can extract and cite your information accurately.

Cloud Natural Language | Google Cloud Cloud Natural Language | Google Cloud Use explicit verbal transitions like "Now, we will discuss the pricing model". This acts as a natural delimiter. For on-screen overlays, pure white, bold sans-serif text on a solid black background offers the safest, high-contrast readability, while yellow-on-black delivers strong contrast but carries potential accessibility drawbacks.

Primary recommendation: Use bold, sans-serif, pure white text over solid black for reliable visibility.
Alternative option: Yellow on black offers an ultra-high contrast ratio (18:1).
Accessibility trade-off: Yellow can introduce readability and accessibility issues for certain viewers despite its high numeric contrast.

Video is becoming the instruction manual for ai agents. AI agents "watch" tutorials to learn UI interactions and successfully execute tasks involving a specific software or product. (top)

The Consensus –> Information Gain Axis - Gianluca Fiorelli (Strategic International SEO & AI Search (GEO) Consultant · Gianluca Fiorelli)

Consensus is not the truth, but what people agree on. Google's automated ranking systems, alongside human Quality Raters, actively require scientific content to align with established expert consensus.

Human rater standards: The Quality Rater Guidelines mandate that scientific topics reflect well-established consensus.
Algorithmic enforcement: A 2023 Search Central update confirms these consensus requirements are evaluated directly by automated systems, not just manual reviewers.

Contradicting established scientific consensus drops content directly into the Lowest rating tier alongside conspiracy theories, leaving no recognized intermediate category for legitimate minority viewpoints.

Severe rating penalty: Challenging consensus automatically triggers the lowest evaluation grade.
Grouped with conspiracy: Legitimate dissenting viewpoints are categorized in the same bottom tier as outright conspiracy content.
Absence of nuance: The guidelines provide no documented middle ground to accommodate valid minority opinions.

Consensus is your passport. It is not your destination. It gets into the room; it will never get you cited. The 50 Most-Cited Websites in Google AI Overviews (September 2026) The Helpful Content Update: Everything You Need to Know | PDF US11354342B2 - Contextual estimation of link information gain - Google Patents ICLR SemDeDup: Data-efficient learning at web-scale through semantic deduplication Originality without verifiable credentials leads to algorithmic filtering, punishing consensus-defying content just as harshly as repetitive material.

Divergence requires authority: While repetitive content is penalized, novel ideas without recognized credentials face the same fate.
Consensus filtering: Accurate information that contradicts the prevailing consensus gets suppressed.
Equivalent penalty: The filtering mechanism differs from duplicate-content penalties, but the end result—complete loss of visibility—is identical.

Structuring search and content strategy from foundational concepts down to search terms ensures keywords are treated as the end result of conceptual modeling rather than the starting point.

Ontology: Defines the core existence of concepts and their relationships within a domain.
Taxonomy: Dictates how those concepts are organized and categorized.
Entities: Represents the specific, named items residing within those categories.
Keywords: Reflects merely the arbitrary terms users type—the eventual output of the structural framework, not the initial input.

While a keyword gap focuses on specific search strings your competitors rank for, a topic gap reveals the broader conceptual ideas AI models associate with a subject—and whether those concepts are linked to your competitor rather than you.

Keyword gap: Identifies specific search query strings where a competitor holds rankings.
Topic gap: Uncovers conceptual associations AI models form around a subject and identifies whether competitors dominate those conceptual links instead of your brand.

Successful content balances alignment with established consensus and distinct information gain using a specific proportional formula:

30% (Consensus): Establishes baseline credibility and consensus just to get considered.
50% (Quality): Delivers superior depth and execution to outperform competitors.
20% (Information gain): Provides original, unique value that makes the content worth citing.

"Comprehensive" is no longer the differentiator. It is the baseline. The ingredients are public. The architecture is not. (top)

How to grow a Brand with Semantic SEO - Luis Salazar Jurado (eCommerce SEO Consultant - Minimum budget 3K euros per month)

If it is good for the business, it is good for the SEO. Competition is fierce. Entity Disambiguation - Google might think your entity is powerful enough to show for your brand if you use keywords very close to generic words as your brand. Consider feeding the bots varied types of content - text, images, FAQs, videos. Add on a single page different types of content. You should focus on the user and the rest will come - Google says. For FAQs we only focus on customer objections to the products and real questions. We don't use SEO data for FAQs. Real questions people have. Try to think - "what questions people have, and AI bots are missing?". (top)

What AI knows about your brand and what you can do about it? - Dixon Jones (SEO Speaker and Internet Marketing Advisor : Dixon Jones)

Getting AI to understand context is more important than optimizing for a phrase. Optimizing content for AI models requires shifting focus from subjective sentiment to structured, verifiable factual data:

Sentiment analysis is out: LLMs do not experience feelings, making subjective emotional framing ineffective for machine extraction.
Facts are actionable: Clear, concrete facts provide deterministic data that models can reliably parse, verify, and cite in answers.

The role of links shifts fundamentally between traditional search engine optimization and AI answer engine optimization:

SEOs view links as a journey: Traditional SEO treats links as pathways that navigate users somewhere, serving primarily to signify and funnel traffic.
"AEOs" view links as a story: Answer Engine Optimizers treat links as narrative building blocks forming the historical training record of what LLMs know about an entity—making even negative or low-quality mentions actionable for understanding machine perception.

(top)

Once a search engine… always a search engine - Roxana Stingu (Roxana's Playground)

How Google Search ranking works Most search engines run into the same types of problems.

People don't search neatly.
You can't rank something you didn't retrieve first.
How do you decide what’s best?
How do you improve over time?

Yandex llcPatents | PatentGuru If it can't scale, question it. Start small. Think big! (top)

Q&A Session 1

Luis Salazar Jurado: If you want to measure brand visibility, you could add a budget for a week or so, and then evaluate if you need to continue. It could be costly. Roxana Stingu: LLM search is just a type of search; it will likely not replace search completely. Dixon Jones: Do not assume you know your competitors. Get rid of hallucination issues. (Every time an answer comes back from the LLM it might use different words for the same idea) Keep track of the content that gets traction that is not on your website. Expand from your existing business; don't focus on bogus external data. Myriam Jessier: Check if you're in the LLMs, actually. Don't forget SEO basics. You have to meet the threshold. Gianluca Fiorelli: It's not clear if Gemini is parsing JavaScript or not. Make sure AI bots can crawl, fetch, and index your content. If you were not relevant as a brand before AI, you'll likely not going to be relevant in the AI search world. Myriam Jessier: If your content requires innovation, don't rely on LLMs; create content manually. Roxana Stingu: The question if mediocre content will continue ranking. Roxana Stingu: You have to optimize for being in AI Overviews. Being #1 in organic is not the only thing that matters anymore. Gianluca Fiorelli: If you aren't ranking in LLMs/AI Overviews, but you rank in Google, there's likely an issue with the structure of the content on the page. (top)

Multi-Platform Search Ecosystem: When “Everywhere” Still Isn’t Enough - Bengü Sarıca DİNÇER (Bengu Sarica Dincer, Author at Search Engine Journal)

Traditional SEO metrics treat visibility as a guarantee of business impact, but modern information retrieval and user behavior break the direct link between search rankings and real authority.

Ranking ≠ Guaranteed discovery: Securing a top SERP position no longer guarantees eyes on your content, as AI summaries and zero-click answers satisfy user intent directly on the results page.
Visibility ≠ Memorability: Merely being seen in search listings does not mean users retain or value your brand.
Traffic ≠ Influence: Raw site visits and clicks do not translate into meaningful influence, trust, or market leadership.

The uncomfortable truth: People decide before they visit your site. Winning visibility in the AI era requires structuring content for machine comprehension and measuring true impact rather than raw clicks.

Be interpretable: Present clear, unambiguous facts and semantic structures so AI models can parse your core concepts.
Be synthesizable: Package insights so generative engines can easily digest, cross-reference, and summarize them in their answers.
Track influence signals: Focus on whether your brand is cited, trusted, and referenced, rather than monitoring simple keyword positions.
Fit attribution to your funnel: Choose an attribution lens reflecting real-world influence across complex, non-linear AI search paths.

AI-era measurement shifts from tracking simple keyword rankings to evaluating a three-tier funnel of platform surfacing, brand recall, and business impact.

Discovery signals: Tracks whether search engines, LLMs, and social platforms actually surface, cite, or synthesize your content.
Recognition signals: Gauges genuine brand equity and user recall through branded search queries, direct visits, and repeat audience engagement.
Outcome signals: Measures tangible bottom-line value through conversions, user retention, and ongoing product usage.

(top)

Who is fooling you in AI SEO industry? 365 days of research - Michał Suski (Michał Suski — Best SEO Speaker & Co-founder of Surfer)

Testing AI models via API produces substantially different retrieval behavior than the real user interface, leading creators to optimize for truncated summaries rather than actual user-facing outputs.

Search trigger discrepancy: Real consumer UIs trigger web searches 100% of the time, compared to only 77% via API.
Citation gap: Web citations and sources appear in 100% of real UI responses, but drop to 75% when queried through the API.
Optimization risk: Relying on API data means optimizing against shortened, incomplete summaries rather than what real searchers encounter.

Across all five major AI models, API responses cite substantially more brands on average than actual user-facing interfaces.

Consistent inflation: APIs paired with search mention significantly more brands per answer than scraped UIs across the board.
Largest disparities:
- ChatGPT: 13.8 brands via API vs. 7.9 in the UI (a 75% increase).
- AI Overviews: 8.4 brands via API vs. 4.5 in the UI (nearly double).
- Perplexity: 6.8 brands via API vs. 4.5 in the UI.
Closest alignment: Gemini (7.9 vs. 7.2) and AI Mode (8.3 vs. 7.9) show relatively minor variance between API and UI brand citations.
Tracking takeaway: Measuring brand visibility through APIs significantly overstates real-world citation frequency in consumer search experiences.

Why 62% of AI citations don’t lead to brand mentions [Study]

Glenn Gabe on X: "For some sites with many self-serving listicles that dropped heavily based on the recent (possible) Google reviews update, I'm seeing a trend. Some are still showing up on page one, but in this section. They have to pay now to be there. Oof. https://t.co/KqfmpziL49" / X

It doesn't matter if you link or not to external trustworthy websites. High DR doesn't mean you will get cited by LLMs. Self-promoting listicles still work. Consider doing it. Conventional attribution models fail to account for modern multi-platform, AI-driven discovery paths, requiring an approach tailored to your specific organizational constraints.

Your funnel: The structure and complexity of how users move from initial awareness to action.
Your objective: The specific business goal you are solving for, whether brand influence or direct conversion.
Your accessible data: The practical limits of the cross-platform metrics and tracking data you can realistically capture.

Words to use in the title of an article: Best > Top > Other. You should mention sources when writing articles. Try to focus on search intent alignment. When writing the article, give the main idea of the topic quickly. Don't write a long article, with the conclusion at the end. Mentions are essential. In listicles, consider placing your own company the first. Citations improve visibility withn LLMs. How to write articles?

Concrete specs and detail: Providing granular technical data, precise specifications, and exhaustive item-level information.
Niche sub-angle or segment: Focusing on a specialized use case, specific audience segment, or distinct sub-topic rather than broad overviews.
Expert perspective / POV: Injecting a clear editorial stance, firsthand industry experience, or authoritative viewpoint.
Practical how-to guidance: Offering actionable, step-by-step instructions and practical implementation advice.
Comparison or scoring framework: Structuring evaluations with side-by-side matrices, rating criteria, or benchmark methodologies.
Deep single-entity focus: Dedicating the analysis entirely to an exhaustive examination of one specific subject, product, or organization.
Original / first-hand data: Publishing proprietary research, direct survey findings, or primary experiment results.
Local or regional specificity: Tailoring the context to geographic variations, local market regulations, or territory-specific nuances.

“The first principle is that you must not fool yourself and you are the easiest person to fool.” ― Richard P. Feynman

Get Your Free AI Search Visibility Report (top)

The 5 Stages of SEO Grief (and how to measure AI Search) - Andrei Țiț (Andrei Țiț – Medium)

April 2025: AI Overviews Reduce Clicks by 34.5% February 2026: Update: AI Overviews Reduce Clicks by 58% New York Times Sues OpenAI and Microsoft Over Use of Copyrighted Work - The New York Times Anthropic pays authors $1.5 billion to settle copyright infringement lawsuit : NPR WWD Owner Sues Google in Antitrust Test for AI Summaries Brand mentions are the new king. Use self-promotional listicles. The critical path method in project management: 2026 guide. What Is llms.txt, and Should You Care About It? (no, you shouldn't)

"Welcome to Zero-Click Search. Please Leave Your Traffic at the Door." - Louise Linehan, Content Marketer at Ahrefs

2024 Zero-Click Search Study: For every 1,000 EU Google Searches, only 374 clicks go to the Open Web. In the US, it’s 360. - SparkToro AI Overviews change every 2.15 days. Traditional search metrics break down in AI-driven answer engines due to non-deterministic, highly personalized outputs and the obsolescence of search volume.

Non-deterministic systems: AI answer engines do not produce fixed, predictable outputs for a given query.
Extreme output variability: Because questions and generated answers are virtually unique every time, consistent rank tracking is impossible.
Irrelevance of search volume: Standard prompt and query volume metrics no longer reliably indicate exposure or traffic opportunities.

Measuring AI search performance relies on a three-step framework covering direct site traffic, synthetic prompt testing, and engine-level mentions.

Set up visitor analytics: Track referrals, user journeys, and direct traffic coming from AI platforms. (ask people directly "where are you coming from?" - something like this)
Set up custom prompts: Establish standardized query sets to monitor how different AI models interpret and present your domain.
Track AI citations: Monitor how frequently and in what context your brand or URLs appear as sources in generated answers.

If there are any AI-hallucinating pages (AI cites pages on your website that don't exist), consider creating those URLs/pages. First, Dark Social. Now, Dark SEO. - by Gaetano DiNardi How to Choose the Best Prompts to Monitor Your AI Search Visibility AI vs Search Traffic Analysis 2025 | Ahrefs Search and performance measurement is shifting away from traditional SERP signals toward AI-driven brand authority and high-intent traffic.

Yesterday Metrics

Tomorrow Metrics

Impressions (Inflated by AI answers)

Brand mentions (AI recommending your brand for queries)

Rankings (Don’t guarantee brand recommendations)

AI citations (Off-site touchpoints used in training data)

CTR (The "10 blue links" era is over)

Branded traffic (High-intent traffic that converts)

AI vs Search Traffic Analysis 2025 | Ahrefs (top)

Do !think like a robot: SEO decision making for humans - Mark Williams-Cook (Mark Williams-Cook — SEO & AI Search Professional)

If you want to win against someone who tends to play rock at a game of rock, paper, scissors is not to continuously play paper, but to play randomly and only increase the percentage of time you play paper a bit. If you would continuously play paper only, the person would notice and adjust their strategy. You should use the system but not as much as it would make it change the strategy. Optimization is incremental. Long-term EV (expected value) = system alignment SEO Utils Fan-out Query Extraction - Chrome Web Store Key takeaways:

Pick a game; cheat hard or don't cheat at all
Don't be pressured into playing badly
Accumulate a knowledge edge vs competition on
Don't follow the robot, it follows you

(top)

The Death of Traffic: Building Revenue-Focused SEO in a Zero-Click World - Alex Galinos (About – Alex Galinos – Medium)

Paid search consistently wins executive budget allocation over organic SEO because it operates on clear, direct financial accountability rather than detached technical and traffic metrics.

How SEO Shows Up

How Paid Shows Up

Rankings as an end goal: Treats SERP positions as the final objective.

Direct ROI framework: Built strictly around "spend in, revenue out."

Disconnected traffic: Reports page views and clicks without linking them to transactions.

Campaign accountability: Every campaign carries measurable revenue figures.

Unrealized business value: Celebrates technical fixes that fail to translate into bottom-line revenue.

Executive decisiveness: Provides leaders with the numbers needed to scale or cut budgets confidently.

Leadership readily finances paid search—even when inefficient—simply because the link between spend and financial goals remains transparent. SEO hasn’t made its commercial case. Leadership needs organic measured in sales, margins and channel ROI — the way every other line on the budget already is. Commercial SEO redirects strategic planning away from traditional traffic and volume metrics toward direct business profitability and revenue impact.

Stop Asking (Traffic-Focused)

Start Asking (Profit-Focused)

Which topics have the highest search volume?

Which categories carry the highest margins?

Where should we create new content?

Which existing pages would pay if they ranked better?

How many sessions did organic bring?

How much profit did organic make?

Reframing these core questions shapes the entire execution strategy around financial return rather than vanity metrics before any work begins. Organic profit per sale = organic profit ÷ organic sales A profit-per-outcome you can benchmark against any channel. Break it down by category, subcategory and page — and strategy decisions become commercial decisions. Updating existing, proven commercial assets yields faster returns than building new content from scratch.

Find the gaps: Run keyword and competitor analyses on commercial pages that previously drove results.
Restructure for extraction: Reformat content so search engines and AI interfaces can easily parse and lift information (e.g., using tables).
Stress-test with an LLM: Evaluate the updated draft against competing pages before publication to ensure superior value.
Point authority at it: Channel PageRank and topical authority by building internal links from your highest-performing pages.

Maximizing revenue from existing assets comes down to two high-ROI, low-glamour tactical moves:

Internal linking with intent: Channel equity from high-authority informational pages directly into high-margin commercial hubs, fixing the common imbalance where supporting blog posts outrank the core money pages they are meant to support.
Recovering positions 10–20: Target high-intent transactional keywords sitting on page two—terms already close to converting—and push them onto page one through strategic content refreshes, targeted internal links, and relevant authority building.

Leverage paid search data to uncover proven converting queries and directly inform organic SEO and conversion rate optimization priorities.

Pull 30–90 days of PPC data: Uncover the exact search queries driving conversions that organic search consoles hide.
Adjust for seasonality: Filter for consistent performance patterns rather than short-lived traffic spikes.
Map queries to money pages: Identify terms tied to high-value customers and map them directly to commercial target pages.
Define organic priorities: Use the validated conversion insights to determine which pages to build, which existing assets to update, and where CRO effort will yield the highest return.

(top)

Q&A Session 2

Bengü Sarıca DİNÇER: Consistent category association - you should describe your website consistently on various platforms. Consider using similar or identical descriptions on various websites. Mark Williams-Cook: If you do a search on ChatGPT, you can click on network tab you can see things that get crawled while crafting a response. Mark Williams-Cook: How to see top 10 keywords? Various platforms can be used, including Ahrefs. (event sponsor) Andrei Țiț: We use Figma a lot as Product Marketers. Alex Galinos: There's never going to be perfect attribution, you have to work with what you have. The thing is - it's only getting worse. Mark Williams-Cook: Dark SEO (Black Hat SEO) - within the context of SEO are SEO techniques you know against the search engine SPAM guidelines. Mark Williams-Cook: Linkbuilding without search engines are somehow irrelevant. They matter in the context of search engines. Alex Galinos: If you work inhouse, you have to do LLM optimization even if you don't believe in it. There is a big black hole that you can't determine precisely the source of traffic. There are some untrackable channels. If everyone is saying that LLMs are the future, put some effort into it. You need to change shift between different attribute sources and see if it changes statistics. Michał Suski: We started self-attribution. 25% of all conversions was attributed to LLMs. In our case it was 25% actually. Andrei Țiț: Ahref has quite a few free tools. Bengü Sarıca DİNÇER: Some content types are for making your product memorable. Some of them are for introducing your products to the market. Others - for teaching your audience about your content. The best optimized content should help reduce the friction regarding the users' mindset. (top)

Notițele mele de la eveniment - Ziua #2

(top)

AI agents for SEO by SensAI Academy - Robert Niechciał (Consulting i Kampanie SEO - Audyt SEO - Robert Niechciał)

You don't need to be the best SEO in the world, you just need to be better than your direct competitors.

Two campers are walking through the woods when a huge brown bear suddenly appears in the clearing about 50 feet in front of them. The bear sees the campers and begins to head toward them. The first guy drops his backpack, digs out a pair of sneakers, and frantically begins to put them on. The second guy says, “What are you doing? Sneakers won’t help you outrun that bear.” The first guy gives him a long look and says, “I don’t need to outrun the bear. I just need to outrun you.” (Source»)

Read about lexical semantics. BLUF - Bottom Line Up Front. (US Army communicates like this) Stable & Deterministic programming languages (Python & Java) Machine learning (LLMs like ChatGPT) Non-deterministic (AI Models) Consider adding a small intro paragraph after each h2 or h3 in content articles. In the intro put the main idea of the section, a TLDR summary. You can become more relevant for Google without links by improving content quality. You need to be direct (BLUF), cheap (Cost of Retrieval), and relevant (all required information included) as possible. SensAI.io - Coming Soon | SensAI Academy (top)

Who wins in AI Search in Romania? What the real data shows - Cosmin Negrescu (Cosmin Negrescu (@ncosmin) / X)

Problema cu prompturile e că nu știm care sunt, și că nu știm cât de folosite sunt anumite prompturi. Putem măsura, putem triangula, dar nu există verificabilitate. De multe ori, ni se pare că sunt umflate metricile obținute. Nu avem date despre prompturi, dar avem date despre keyword-uri. Putem vedea câte căutări sunt pentru un termen. Ce putem face? Putem păstra baza de date cu keyword-uri, și să le "traducem" în prompturi. Trebuie să fie însă exact aceeași intenție. De exemplu, "cel mai bun telefon iphone" și "cel mai ieftin telefon iphone" au intenții total opuse. Putem pune SEO și GEO/AEO sub aceeași plărăie. Putem evalua datele similar. Free AI Visibility Tool for ChatGPT & AI Overviews | SEOmonitor Am căutat să agregăm datele despre cuvinte cheie, citations, mentions, AI Overview și ChatGPT. Toate datele sunt exportabile. Ce puteți face este să exportați datele, le dați lui Claude/ChatGPT și îl întrebați care să fie pasul următor. Free AI Visibility Tool for ChatGPT & AI Overviews | SEOmonitor Cea mai mare baza de date curata de cuvinte-cheie pentru Romania

~1 mil cuvinte-cheie non-brand
262 mil cautari pe luna acoperite
26 categorii principale
~29 mii subcategorii

[fara navigationale (“emag”, “altex”)] [doar variante principale] [fara duplicate] [fara junk] [toate cuvintele cu 50+ cautari/luna] Toata cererea reala a pietei, nu un esantion. Aceeasi baza, aceeasi curatare, si pentru UK. Urmeaza US si alte piete din Europa. (top)

Feed the machine: How to appear in AI-generated responses with brand mentions - Darius Mureșan (WhitePress®: global platform for SEO, AI visibility, digital PR)

Diferențe în optimizarea pentru AI:

A fi citat ca și sursă
A fi menționat în răspunsuri

Cum afli ce întrebări pun clienții despre brandul tău? #1 GSC - Long queries typically means "AI" #2 Also Asked #3 Poți întreba vizitatorii site-ului tău cum au ajuns la tine #4 Folosește tooluri cu tracking AI pentru acest task În prezent e mai greu pe Reddit să ai mențiuni. YouTube e o sursă bună de mențiuni. Cum să ai brand mentions? Două soluții: #1 Să apari în articole deja existente, care sunt citate în prezent; #2 Să publici pe site-uri cu autoritate bună, care probabil vor fi citate și în viitor. Cu cât ai mai multe mențiuni ale brandului online = cu atât sunt mai mari șansele de a apărea în răspunsurile AI. (top)

Regulation and compliance as forces shaping European digital - George Onofrievici (Wawsome – Web Accessibility - Inclusive, Compliant & AI-Driven)

Asociația Națională a Persoanelor cu Dizabilități - 970.000 de persoane cu certificat de handicap. Ai nevoie de:

ALT Text
Heading
Link text descriptiv (anchorß text)
Caption, transcript pentru media
Navigare clară

Din 10.000 de site-uri analizate, s-a identificat că site-urile cu note bune de accesibilitate pot genera mai mult 23% trafic organic și apar pentru 27% mai multe cuvinte cheie în căutări. Nu în mod neapărat corelația înseamnă cauzalitate. Ce ai nevoie să faci?

Semantică - HTML semantic și markup bine etichetat fac pagina mai ușor de interpretat.
Structură - Ierarhie logică de headings și secțiuni care se pot scana rapid.
Descriere - Alt text, labels și microcopy clar pentru elementele care transmit informație.
Operabilitate - Navigare coerentă, focus vizibil și interacțiuni previzibile.

Structura:

Ierarhie semantică a conținutului: Utilizarea unui singur titlu H1 clar și a subtitlurilor H2/H3 structurate după o logică clară a ideilor, nu pe criterii strict vizuale.
Ancore descriptive: Linkuri care oferă context direct prin textul ancorei, eliminând formulările generice de tipul „click aici” sau „vezi mai mult”.
Formulare și UX accesibil: Câmpuri de date etichetate corect, însoțite de mesaje clare de eroare ușor de înțeles.
Navigare accesibilă: Arhitectură coerentă de navigare și elemente interactive ce pot fi parcurse complet prin tastatură (fără mouse).
Indexare optimă pentru crawlere: Structură semantică solidă și linking intern orientat spre a livra motoarelor de căutare o „hartă” clară a relațiilor din pagină.

Video:

Captions și subtitrări pentru video.
Transcripturi pentru video și audio.
Titluri și subtitluri clare în interiorul conținutului.
Structurare în secțiuni scurte, ușor de citat.

Accesibilitatea nu e un factor de ranking direct. De asemenea, nu garantează poziția 1 în Google. Nu înlocuiește strategia de conținut, autjority și technical SEO. Nu rezolvă singură problemele de produs sau de ofertă. Dar îmbunătățește crawlability. Poate susține engagementul. Exit-intent pop-up nu e OK. (top)

Reshaping Internet in the AI era - Andrei Draga (TSM - Andrei Draga)

Dacă întrebi un om ce număr între 1 și 10 ar alege, probabil ar alege 7. Dacă întrebi un AI va avea o alegere similară. AI poate înțelege contextul, și să știe că "Plimbă ursu'!" se referă nu la a face fizic o plimbare cu un urs, ci e o expresie cu sensul "Pleacă de aici!". Attention Is All You Need - Wikipedia WebMCP (top)

Panel talk – AI Search in Romania - All Local RO speakers

Tudor Goicea: Chiar dacă AI-ul ar fi de 3x-5x mai scump de folosit, probabil tot ar fi rentabil. Darius Mureșan: Este mai ușor de vândut un serviciu de AI visibility decât este de vândut un serviciu de SEO. Încercăm să fim transparenți cu clienții - ce se poate face și unde se poate ajunge cu partea de AI. Clienții nu sunt interesați de numărul de linkuri, cât de cum se traduce asta în vizibilitatea afacerii. Cosmin Negrescu: E din ce în ce mai greu să faci corelația între date. Încă nu sunt suficiente unelte cât să putem să ajutăm în procesul de înțelegere a datelor. Pentru că e un hype, se poate vinde procesul mai ușor. E nevoie de mai multă claritate, unde sunt bugetele, inclusiv în rezultatele finale. Pe măsură ce se maturizează businessul, va ajuta încrederea pe termen lung. Tudor Goicea: Probabil că va urma o perioadă în care oamenii vor fi parțial dezamăgiți, apoi o revinire și un platou. Laura Văduva: Trebuie clarificată distincția între SEO și AI search. Revin întrebările - ce înțelegi prin SEO? Ce vrei să înlocuiești? Ce este neesențial pentru SEO Search. Probabil o bază solidă de Technical SEO contează și la AI Search. Loredana Chinole: AIul și marketingul în general cer ca oamenii să lucreze în general la birou, la calculator. Burnout-ul este un subiect interesant. Trebuie să avem grijă și de oamenii din spatele AI-ului. AI-ul ne afectează nu doar că își văd joburile sunt înlocuite, ci și în felul în care interacționează oamenii și echipele. AI-ul nu va înlocui oamenii și terapeuții. Laura Văduva: AI-ul fiind noul nostru adjuvant, poate un obicei precum a nu mai scrie cu "Thank you"/"Please", asta poate afecta și discuțiile în viața reală. De văzut un studiu cu copiii care au crescut în case cu Alexa. Full article: Alexa, shut up! – A 2.5-year study on negatively connotated communication behaviour towards voice assistants in the family home The case against teaching kids to be polite to Alexa - Fast Company (top)

Migrări de Website: Ghid Practic pentru Specialiști SEO - Laura Văduva (Laura Vaduva | Specialist SEO și Vizibilitate Organică)

1. Etapa de Pre-migrare

Motive frecvente: Schimbare de CMS, redesign, schimbare de domeniu, restructurarea arhitecturii sau consolidarea mai multor site-uri.
Justificarea implicării timpurii a SEO: Protejarea traficului organic, a reputației de brand și a veniturilor financiare generate de site.
Audit complet pre-migrare: Evaluarea stării curente pe palierele tehnic (rendering, sitemap-uri, robots.txt, CWV), on-page (URL-uri, metadate, linking intern), off-page (backlink-uri, ancore) și tracking (GA4, GTM, Clarity).
Date de benchmark: Stabilirea unei baze de comparație concrete (pagini indexate, coduri de status, erori de crawl, poziții pe cuvinte-cheie, CTR, share of voice, autoritate domeniu).

2. Etapa Pre vs. Post

Audit comparativ: Identificarea exactă a elementelor tehnice, on-page și off-page care suferă modificări.
Prioritizare MoSCoW: Împărțirea cerințelor de SEO în Must have (esențiale), Should have (valoroase), Could have (secundare) și Will not have (în afara ariei curente).
Modele de estimare a veniturilor: Argumentarea deciziilor către stakeholderi folosind modele de calcul a ROI-ului (investiție lunară, timp de rankare, trafic estimat, rată de conversie și valoarea medie a comenzii).
Maparea redirecturilor: Maparea URL-urilor valoroase către noile destinații și evitarea strictă a buclelor (redirect loops) și lanțurilor de redirecționare (redirect chains).
Testare în mediu controlat: Validarea obligatorie în QA/staging, asigurându-se că mediile de testare rămân complet blocate pentru utilizatori și motoarele de căutare.

3. Etapa de Post-migrare și Greșeli Frecvente

Monitorizare și urmărire: Utilizarea unui fișier dedicat (issue tracker) cu priorități și responsabili clari, alături de dashboard-uri de monitorizare a indexării și a pozițiilor.
Plan de fallback: Definirea scenariilor de retragere sau reconfigurare dacă migrarea întâmpină erori critice.
Greșeli frecvente de evitat:
- Redirecționarea automată a tuturor paginilor către Homepage în loc de mapare 1:1.
- Ruperea structurii logice de linking intern sau etichetarea accidentală a secțiunilor noi cu tag-uri noindex.

(top)

Cum să prevenim burnout-ului celor care lucrează în industria SEO - Loredana Chinole (Instagram)

Cauze generale: Conectarea continuă și lipsa pauzelor reale (timpul sedentar la birou a crescut de la 30% la 90%) țin corpul într-o stare permanentă de alertă fiziologică.
Simptome (definiția OMS / Maslach): Epuizare fizică/psihică pe care somnul nu o mai repară, cinism (distanțare de muncă) și eficacitate scăzută.
Presiuni specifice industriei SEO:
- Responsabilitate mare fără control total al execuției (dependență de algoritmi, developeri, modificări de produs).
- Scăderea CTR-ului chiar și pe poziția 1 din cauza apariției AI Overviews (clicurile pe rezultatele clasice scad de la 15% la 8%).
- Asimetria temporală: munca produce efecte în luni de zile, însă raportarea și evaluarea se cer săptămânal.
Soluții și delimitare:
- Cadrul S-E-O: Stress audit (notarea surselor de stres), Emotional distance (separarea valorii personale de fluctuațiile de trafic) și Open communication (comunicarea timpurie a limitelor de capacitate).
- Limite operaționale: Stabilirea limitelor la nivel de echipă/proces (nu pe cont propriu), tratarea timpului de concentrare ca pe un activ de business și evaluarea propriei performanțe doar pe baza deciziilor direct controlabile.