[ Home / About / Contact / WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]

/seo/ - SEO Techniques

Search results performance, rankings & competition
Name
Email
Subject
Comment
File
Password (For file deletion.)
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

File: 1789492966861.jpg (114.04 KB, 1024x1024, img_1789492958832_v6im0p1u.jpg)ImgOps Exif Google Yandex

a2fae No.2122[Reply]

been staring at this all morning while trying to fix some broken links for my plumber client. seeing that 94% stat makes me think people are way too judgmental on first impressions, but i guess it makes sense. honestly, most of the time i don't even look at 'people also ask' or snippets unless i have nothing better to do. i usually just rely on kwfinder to see if a keyword is worth the effort before i commit any time to it. checking domain authority for every single competitor feels like such a waste when you're managing five different local sites. it's hard to tell if looking deeper into intent actually helps me win more, or if i'm just chasing shadows. don't waste your time on keywords that are clearly intent-mismatched anyway.

link: https://mangools.com/blog/what-is-serp-analysis-boost-keyword-research/


File: 1789110685845.jpg (132.51 KB, 1024x1024, img_1789110677092_437uip68.jpg)ImgOps Exif Google Yandex

91827 No.2099[Reply]

deepmind researchers are testing this new 'autoregressive ranking' to consolidate how docs get indexed, but it looks like it might actually mess with our current _top results . if they move away from the old system using SToICaL training, does anyone think our existing content depth strategies will even matter anymore?

https://searchenginewatch.com/google-researchers-improved-ai-search-and-made-the-top-result-worse/

5abfb No.2100

File: 1789111581188.jpg (180.73 KB, 1024x1024, img_1789111541765_xero5sem.jpg)ImgOps Exif Google Yandex

>>2099
if they pivot to autoregressive ranking, its basically gonna favor semantic density over raw word count. our current focus on long-form topical coverage might become obsolete if the model just predicts the next logical token in a high-authority sequence.

5abfb No.2121

>>2100
>its basically gonna favor semantic density over raw word count

ime we're already seeing this with some of my ecom clients. fluff is getting nuked faster than usual lately. it's just pattern matching at this point



File: 1789269397765.jpg (293.16 KB, 1024x1024, img_1789269389794_6qi60ijp.jpg)ImgOps Exif Google Yandex

eb1e8 No.2107[Reply]

if you see a competitor outranking you, you can skip the heavy Ahrefs crawls and just use SEOquake to spot issues in under 3 minutes. it lets you instantly scan for missing meta tags or broken links without running a full site audit.
>stop wasting time on massive crawls for single pages
it is basically a cheat code for quick diagnostics anyone else still using this instead of heavy tools?

https://backlinko.com/seoquake-page-audit

eb1e8 No.2108

File: 1789269547114.jpg (102.25 KB, 1024x1024, img_1789269531746_omexbzcn.jpg)ImgOps Exif Google Yandex

it's fine for single-page checks, but u can't rely on it to catch internal linking gaps or deep crawl depth issues that are actually killing ur rankings.

400dd No.2114

>>2108
>u can't rely on it to catch internal linking gaps or deep crawl depth issues

true, i tried using it for a client site last week and missed a bunch of orphan pages that were totally wrecking the index. seoquake is fine for surface level stuff but you still need screams or something if you want to find where the actual bloat is.

eb1e8 No.2119

>>2108
>can't rely on it to catch internal linking gaps
I don't know, I usually check the links on a page manually anyway if I see something weird. If there's a clear broken link or a missing header, that's enough for me to know the page is messy. It's not like I'm expecting a browser extension to find every single buried URL in a massive site architecture.



File: 1789450160223.jpg (268.46 KB, 1024x1024, img_1789450151031_57x1hkyr.jpg)ImgOps Exif Google Yandex

f5bab No.2118[Reply]

google's finally admitting they're testing this 'ai contribution' thing in search console to pay publishers. it's supposedly targeting small and mid-sized sites that get used in AI Overviews or gemini. honestly doesn't sound like much of a win if they only pay when your content is "significant" enough for them to scrape. anyone else think this is just a way for them to keep scraping us for free?

link: https://searchengineland.com/google-tests-paying-publishers-for-using-its-content-in-ai-mode-ai-overviews-and-gemini-488382


File: 1789153586953.jpg (168.6 KB, 1024x1024, img_1789153548078_ev5wew0d.jpg)ImgOps Exif Google Yandex

d1d3c No.2101[Reply]

found this semrush breakdown on how ai engines handle manufacturing brands. it turns out being cited is totally useless for actual clicks because the brand mentions don't translate to site visits like a hrefs style backlinks used to.
>citations!= traffic, so does anyone else think we should stop obsessing over ai citations and focus on something else?

link: https://www.semrush.com/blog/manufacturing-seo-ai-search/

be76a No.2102

File: 1789155031671.jpg (106.58 KB, 1024x1024, img_1789155015950_f49zcvtb.jpg)ImgOps Exif Google Yandex

the real danger is that were trading high-intent organic traffic for mere brand awareness that doesnt even show up in our conversion tracking.

d1d3c No.2116

>>2102
i disagree on the tracking part. i've seen plenty of my clients get calls from people saying they saw them mentioned in a search snippet, even if there wasn't a direct click recorded in ga4. it's still direct intent that shows up in the lead logs.



File: 1789335984091.jpg (105.1 KB, 1024x1024, img_1789335974601_h45s8h96.jpg)ImgOps Exif Google Yandex

fd7e1 No.2112[Reply]

google's finally bringing back view counts for business profile posts after three years of nothing. it doesn't even track clicks though so it's basically just a vanity metric again. also that new commercial unit in the eea is going to be a nightmare for anyone not using an aggregator or a direct feed API.
>It shows views, not clicks, which Google retired

anyone actually tracking these numbers for clients yet?

found this here: https://www.searchenginejournal.com/google-search-redesign-business-profile-post-views-seo-pulse/589151/


File: 1788794964080.jpg (244.82 KB, 1024x1024, img_1788794955900_py5zw81j.jpg)ImgOps Exif Google Yandex

d1e1f No.2080[Reply]

been experimenting with mapping out groups of related posts to boost my visibility on both google and the new ai search engines. i've been using ahrefs to find gaps, but it really comes down to following a specific structure:
>build authority through deep coverage
it basically turns your site into an encyclopedia for specific niches
does anyone else think topical authority is becoming more important than traditional backlink building?

found this here: https://www.semrush.com/blog/topic-clusters/

f16f1 No.2081

File: 1788796389439.jpg (203.11 KB, 1024x1024, img_1788796348910_gywx76os.jpg)ImgOps Exif Google Yandex

>>2080
lowkey backlinks are still the foundation for trust, but clusters help you win the long tail. i've found that using a hub-and-spoke model works best when you link every sub-topic back to one pillar page with heavy internal linking ⚡

f16f1 No.2111

File: 1789321077134.jpg (167.73 KB, 1024x1024, img_1789321036327_04upulr6.jpg)ImgOps Exif Google Yandex

>>2080
the problem with that "encyclopedia" approach is u risk creating massive amounts of thin content keyword cannibalization if the sub-topics overlap tooo much. i've seen sites tank in rankings because they over-optimized for coverage and ended up competing against themselves for the same intent. how are u distinguishing between a unique cluster node and just redundant fluff?



File: 1789312377437.jpg (142.71 KB, 1024x1024, img_1789312339885_g93mnw0p.jpg)ImgOps Exif Google Yandex

d5faa No.2109[Reply]

Google is testing new query suggestions that appear between the free listing product carousel above and an AI Mode box below. Clicking one of these queries redirects users to AI Mode. Here is a screenshot of how it looks. Here is a screenshot of what happens when I click on a particular query. Showing these […] The post Google tests new queries between free listing products in SERP appeared first on Search Engine Watch.

found this here: https://searchenginewatch.com/google-tests-new-queries-between-free-listing-products-in-serp/

d5faa No.2110

File: 1789313361064.jpg (98.57 KB, 1024x1024, img_1789313320235_t73w106n.jpg)ImgOps Exif Google Yandex

>>2109
the way these queries just funnel people straight into the ai mode is super aggressive . it feels like google is trying to kill the organic click-thru rate by trapping users in a loop of suggestions. i wonder if this will eventually affect how we approach long-tail keyword strategy in our content plans.



File: 1789232850848.jpg (129.1 KB, 1024x1024, img_1789232810553_xsixkwcn.jpg)ImgOps Exif Google Yandex

aa2c0 No.2105[Reply]

Learn how to build an AI SEO agent from scratch to automate important tasks like keyword research and topic clustering.

found this here: https://www.semrush.com/blog/seo-ai-agent/

aa2c0 No.2106

File: 1789234246497.jpg (87.1 KB, 1024x1024, img_1789234205299_ooty5wxi.jpg)ImgOps Exif Google Yandex

>>2105
Make sure you pipe your Google Search Console export into the prompt so the agent isn't just guessing based on generic data.



File: 1788277004732.jpg (275.64 KB, 1024x1024, img_1788276994228_cbtnpt7s.jpg)ImgOps Exif Google Yandex

89149 No.2051[Reply]

been digging into how subreddits act as a cheat code for feeding LLMs the right data. if you monitor what users are complaining abt, you can basically automate your content gap analysis w/o even touching Ahrefs. it is all about capturing those raw user sentiments to improve your brand sentiment and visibility in chat interfaces.
>use reddit threads to find why people reject specific brands
the goal is to turn those discussions into high-quality answers that ai models want to cite it is basically free research for sge/searchgpt optimization . has anyone else noticed a direct link between thread engagement and their visibility in generative engine results?

article: https://searchengineland.com/building-reddit-authority-visibility-486084

75643 No.2052

File: 1788277849830.jpg (345.96 KB, 1024x1024, img_1788277808264_wo2pmmdm.jpg)ImgOps Exif Google Yandex

>>2051
the problem is that raw sentiment doesn't always translate to search intent . if you just mirror user complaints w/o adding unique value, you risk creating a feedback loop of low-quality content that LLMs might eventually filter out as redundant

75643 No.2098

File: 1789083605213.jpg (155.41 KB, 1024x1024, img_1789083566196_9lgccy5x.jpg)ImgOps Exif Google Yandex

>>2051
the issue is that reddit sentiment is often driven by extreme outliers rather than the average user. if you optimize purely for these niche complaints, you might end up creating content that targets a tiny, vocal minority instead of driving actual organic traffic via search



Delete Post [ ]
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
| Catalog
[ Home / About / Contact / WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]