Eleven principles, and the tactics we will never automate.
AwesomeSEO is opinionated on purpose. These rules decide what the product recommends, what it measures and what it refuses to do. Where a rule rests on evidence, its tier sits beside it.
The principles
- 01
Evidence over assertion
Every number carries its source and age; every rate an interval and a sample size; every practice an evidence tier. When the evidence is a correlation we say so, and a change that is not significant is marked n.s. instead of celebrated.
106 practices across 7 tiers · See the evidence - 02
The top 20 is the competitive set
Below the top 20 the order is mostly noise; inside it, pages Google already judged relevant compete on quality. For every money keyword we keep a standing study of those 20 results: who ranks, what they cover and what is left to add.
practitionerTreat the top 20 as the competitive set and watch it for years - 03
Top three on many metrics, never an outlier on one
Advantages compound across many signals. We compare your page with the top 20 metric by metric and close the gaps where you sit outside the podium. Page-1 averages are benchmarks, not targets to beat tenfold.
practitionerBe top three on many metrics, never an outlier on one - 04
Entities over keywords
Search engines understand things, not strings. Name the real products, places, people and standards, describe them through their attributes, and the keyword variants follow.
measuredWrite about entities; keywords are their attributesNumber of Distinct Entities Used: in the top 200 for 81 of 93 months, average r −0.27 - 05
Define the source context first
Before the first page, decide who the site is: one central entity and one perspective that frames every page, product and post. Everything we plan is measured against that context, and drift is flagged before it publishes.
documentedDefine the source context before the first page - 06
Publish complete clusters, then expand to adjacent ones
Ship a whole cluster, typically 5 to 20 pages, with its internal links mapped, so the coverage shows in one crawl. Then choose the next cluster by how close it sits to what already ranks.
practitionerPublish complete clusters, not single postspractitionerExpand only to adjacent clusters - 07
Refresh for real
An update means the page changed: at least 10 to 15% of it revised, with facts and years brought current. Changing only the date misleads readers, so the product never does it for you.
documentedRefresh for real: money pages monthly, everything within 90 daysReview cadence: money pages every 30 days, others within 90 - 08
Keep the cost of retrieval low
Fast, clean pages are cheaper to crawl, render and quote. Keep the server response under 200 ms, and under 100 ms on money pages, put the content in the initial HTML, and make structured data match what the page shows.
measuredKeep cost of retrieval low: server response under 200 msLoad Time Milliseconds: #6 of 200, r −0.23 - 09
Grow brand demand with links
Links and branded searches grow together on sites people seek out; many links and few branded searches look bought. We track referring domains per 100 branded searches against your niche, with mentions and reviews beside them.
documentedGrow branded search alongside linksNumber of Referring Domains: #1 of 200, r −0.28 - 10
Agents do the work; a named person approves it
Agents crawl, analyse, draft briefs, write structured data and check feeds on a schedule. Nothing reaches a store, a site or an inbox until a person with the admin role approves it, and outreach goes out only from real people under their own names.
Every approval is logged with a name and a time, and can be undone - 11
Declare AI use
Google accepts AI-assisted content that helps people; scaled low-value content breaks its spam policies. We label how AI helped on every page an agent touched and keep a human review on every draft.
confirmedDeclare AI use and keep a human in the loop
10 tactics the product will never automate.
Some tactics work until the next update. These break Google’s spam policies, consumer law or plain honesty, so AwesomeSEO refuses to automate them and says why when you ask.
Never automated 10
Fake clicks violate Google’s spam policies and click systems filter them.
Hidden text and cloaking are named spam policy violations.
Link spam under Google’s spam policies.
Instead: Mark paid placements rel="sponsored".
Expired domain abuse became a spam policy in March 2024.
Instead: Acquire domains only as real businesses or content you continue.
Manipulative use of annotations; no evidence it works.
Instead: Use hreflang only for real language versions.
Deceptive and against platform rules; worthless as a signal.
Instead: Seed discussion with real people only.
Deceptive, and illegal in some markets.
Instead: Outreach agents draft from real, named senders and need approval to send.
Doorway abuse under Google’s spam policies.
Instead: One page per real service area.
Misleads readers; Google advises against it.
Instead: Update at least 10–15% of the page or leave the date.
Breaks Google review policies and consumer law.
Instead: Collect reviews from real customers only.
Retired practices we flag, never recommend 9
INP replaced FID as a Core Web Vital on 12 March 2024.
Instead: Measure INP instead.
Not Google metrics. The Core Web Vitals are LCP, INP and CLS.
Instead: Ignore.
Google retired the sitelinks search box in November 2024.
Instead: Remove nothing, expect nothing.
HowTo rich results were removed in 2023; FAQ rich results are limited to authoritative government and health sites.
Instead: Keep FAQ content for readers; do not expect the rich result.
Google ignores priority and changefreq and retired the ping endpoint in 2023.
Instead: Use accurate lastmod and Search Console.
Removed in 2022.
Instead: Use canonicals, robots rules and clean internal links.
Sunset in September 2023.
Instead: Use server-side or another testing tool.
Clean keyword density fell 98 places and exact matches in H1–H6 fell 68 places in a year.
Instead: Cover variants and entities instead.
No evidence they affect rankings; HTTPS itself is the confirmed signal.
Instead: Ignore for ranking.
Work with a tool that says no to the wrong things.
Open the demo workspace and follow any recommendation back to its evidence tier and its data.