Serge Bezborodov - AI Internal Linking for E-Commerce Websites

Transcript

1. Serge Bezborodov AI Internal Linking CTO of JetOctopus for e-commerce websites
2. x.com/Sergebezborodov In Ideal World
3. x.com/Sergebezborodov In Real World
4. x.com/Sergebezborodov
5. x.com/Sergebezborodov
6. x.com/Sergebezborodov Old Days: Simhash Link building 0000000000100000010011000010000110001000000110000100001000110000 Technical SEO 1100100000001010010100010011000010001001000100000000001100001100 Hamming distance: Similarity score: 0.6875 20
7. x.com/Sergebezborodov Text Embeddings Link building -0.028526846319437027, -0.009915490634739399, 0.008613178506493568, -0.0028113413136452436, -0.02523316629230976, -0.02877490594983101, -0.04445778205990791, -0.009646759368479252, -0.04605638608336449, 0.02261476032435894, 0.020313318818807602, -0.0032385550439357758, -0.026032470166683197, -0.025894658640027046, 0.01249944418668747, -0.022973068058490753, 0.012823299504816532, -0.02747948281466961, -0.004578765481710434, 0.027327891439199448, 0.04861966893076897, -0.020947249606251717, 0.058376677334308624, 0.03428734093904495, -0.013278075493872166, -0.007834547199308872, 0.04233549162745476, 0.006580468267202377, 0.02891271561384201, 0.004492633510380983, 0.03924852982163429, -0.04870235547423363, 0.043189920485019684, 0.05112782493233681, -0.019334862008690834, 0.025784410536289215, 0.0014470139285549521, -0.01194131001830101, 0.031090127304196358, 0.013243623077869415, 0.04164643958210945, -0.0026580267585814, 0.058542050421237946, 0.0037381192669272423, -0.010273799300193787, 0.022270234301686287, -0.03139331191778183, -0.022008392959833145, 0.050714392215013504, 0.056061454117298126, -0.010570092126727104, -0.015558845363557339, 0.018480435013771057, 0.043548230081796646, -0.018025659024715424, -0.003941390197724104, 0.0164270531386137, 0.0055951206013560295, -0.022215109318494797, 0.03494883328676224, -0.029381273314356804, -0.007641611620783806, 0.000496549706440419 Technical SEO -0.01993323303759098, 0.035309500992298126, -0.006206420715898275, -0.0331009104847908, -0.013747780583798885, -0.007094050291925669, -0.023274077102541924, 0.02904716692864895, -0.0025790194049477577, -0.016732174903154373, 0.05764702335000038, -0.014509604312479496, -0.004053743090480566, -0.03352026268839836, 0.019569793716073036, 0.012951009906828403, -0.013538104481995106, -0.04733094573020935, 0.012356927618384361, 0.002862082328647375, 0.06704052537679672, -0.05683627352118492, 0.014251003973186016, 0.026251481845974922, 0.03248585760593414, 0.033492304384708405, 0.0361761637032032, 0.08767268061637878, 0.024070847779512405, -0.0030298235360533, 0.003337348811328411, -0.04224279895424843, -0.012978967279195786, 0.004979813937097788, 0.035644982010126114, -0.0012659210478886962, 0.0243364367634058, -0.019038613885641098, 0.04090087115764618, -0.018731089308857918, 0.014509604312479496, 0.0010029519908130169, 0.03502993285655975, 0.006202925927937031, -0.012957999482750893, -0.0068039982579648495, -0.0074994247406721115, 0.011231663636863232, 0.02616761066019535, -0.0044730957597494125, 0.05524273216724396, -0.021331075578927994, 0.029606303200125694, 0.08409420400857925, -0.01910850591957569, -0.022239673882722855, -0.039726682007312775, 0.008240281604230404, 0.012440797872841358, 0.054236285388469696, -0.01797625422477722, -0.013607996515929699, 0.010966073721647263, -0.027439646422863007
8. x.com/Sergebezborodov Cosine Similarity Link building Technical SEO 0 1 0.5144
9. x.com/Sergebezborodov Cosine Similarity in deg
10. x.com/Sergebezborodov
11. x.com/Sergebezborodov Text Embeddings Accuracy Link building Technical SEO 0 1 0.5144 ~ 59° Website Internal Linking Technical SEO 0 1 0.4579 ~ 62.8°
12. x.com/Sergebezborodov The Current Manual Approach Page B Page C Page D Page E Page F Page G Page H Page I Page J Page A Candidates
13. x.com/Sergebezborodov The Current Manual Approach Page B Page C Page D Page E Page F Page G Page H Page I Page J Cosine similarity 0.5134 0.8258 0.2791 0.7477 0.2010 0.7123 0.3572 0.6696 0.5915 Candidates Page A
14. x.com/Sergebezborodov The Current Manual Approach Page B Page C Page D Page E Page F Page G Page H Page I Page J Cosine similarity 0.5134 0.8258 0.2791 0.7477 0.2010 0.7123 0.3572 0.6696 0.5915 0.5134 0.8258 0.2791 0.7477 0.2010 0.7123 0.3572 0.6696 0.5915 Best candidate Candidates Page A
15. x.com/Sergebezborodov
16. x.com/Sergebezborodov The Current Manual Approach Page A Page B Page C Page D Page E Page F Page G Page H Page I Page J Page K Page L Page M Page N Page O Page P Page Q Page R Page S Page T Page U Page V Page W Page X Page Y Page Z Page AA Page AB Page AC Page AD Page AE Page AF Page AG Page AH Page AI Page AJ Page AK Page AL Page AM Page AN Page AO Page AP Page AQ Page AR Page AS Page AT Page AU Page AV Page AW Page AX Page AY To check 10k pages we need 100.000.000 calculations
17. x.com/Sergebezborodov Thin Content Problem
18. x.com/Sergebezborodov Thin Content Problem
19. x.com/Sergebezborodov vs ~12 words from product title & description
20. x.com/Sergebezborodov Solutions
21. x.com/Sergebezborodov The Most Popular Don’t do internal linking, leave it as it is
22. x.com/Sergebezborodov Another one Accept low quality, better than nothing
23. x.com/Sergebezborodov Categorisation
24. x.com/Sergebezborodov Split by Categories Levels Page B Page C Page D Page E Page F Page G Page H Page I Page J Same category Page A
25. x.com/Sergebezborodov Split by Categories Levels Page B Page C Page D Page E Page F Page G Page H Page I Page J Parent category Same category Page A Page B Page C Page D Page E Page F Page G Page H Page I Page J
26. x.com/Sergebezborodov Split by Categories Levels Significantly speed up the process Increasing relevancy
27. x.com/Sergebezborodov Implementation
28. x.com/Sergebezborodov Implementation 1. We don’t change existing internal linking 2. We build a layer on top of it 3. We test and experiment
29. x.com/Sergebezborodov Interlinking Core Concepts Donors Acceptors strong pages with crawl budget with impressions weak pages no/low crawl budget
30. x.com/Sergebezborodov Interlinking Core Concepts Donors Acceptors strong pages with crawl budget with impressions weak pages no/low crawl budget links
31. x.com/Sergebezborodov The TODO 1. Choose acceptors pages - what
 do you want to improve - Is Indexable - Yes - Exclude pagination - Exclude pages with GET params - Catalogs with products - Product pages with stock - Orphans or low internal linking - Business value pages
32. x.com/Sergebezborodov The TODO 2. Choose donors pages - Is Indexable - Yes - Preferably with crawl budget
33. x.com/Sergebezborodov The TODO 3. Matching fields - Breadcrumbs - Check your pages have JSON+LD
 breadcrumbs schema - Otherwise create custom extraction
 for each BR level for the URL field
34. x.com/Sergebezborodov The TODO 4. AI Distance (similarity) - 0.6
35. x.com/Sergebezborodov The TODO 5. Anchor generation - simple solution - use target (acceptor) page title or H1 - add some regexp cleaning on top (remove patterns, brand and so on) - better solution - use your keywords data for anchors
36. x.com/Sergebezborodov The TODO 6. That’s it - IL ready to generate
37. x.com/Sergebezborodov Tips & Tricks Generate Asses Finetune Generate 1. Start with small batches about 1-10-20k URLs 2. Play with parameters, it’s iterative process 3. Deploy and test the changes
38. x.com/Sergebezborodov Introducing JOIL
39. x.com/Sergebezborodov
40. x.com/Sergebezborodov What It Can? -free, absolutely free -no limitations, only the capacity of your PC -works natively with Screaming Frog exports -does not duplicate existing internal linking -extracts automatically Breadcrumbs from schema -two linking algorithms -anchor generation -anchor blending by your list (GSC keywords, semantic) -it’s fast and hungry to RAM
41. x.com/Sergebezborodov Download https://github.com/jetoctopus/joil
42. x.com/Sergebezborodov What We Want To Do? OpenSource
43. x.com/Sergebezborodov Thank you! LinkedIn