The old backlink brief was one-dimensional: get a followed link from a strong domain.
That is no longer enough.
A third-party page can now serve at least three different systems:
- A human deciding what to trust.
- A search engine evaluating and ranking documents.
- An AI answer system retrieving sources before composing a response.
The same link can help all three, one of them, or none.
This changes how you should evaluate a backlink opportunity. Domain Rating and anchor text still describe part of the placement. They do not tell you whether the surrounding page contains useful evidence, whether it answers a real buyer question, or whether an answer engine would have a reason to retrieve it.
The source behind this idea
This field note is based on Ethan Smith’s interview on Lenny’s Podcast about answer engine optimization. Smith is the CEO of Graphite and discusses work his team performed for Webflow.
The interview is a practitioner account, not an independent controlled study. Its client figures and observations should be treated as reported experience, not universal benchmarks.
The most useful part is not a performance number. It is the separation between two jobs:
- Build detailed owned content that answers specific questions.
- Appear in the third-party sources that answer engines retrieve and cite.
Smith describes owned content around use cases, features, integrations, and languages. Off-site, the work shifts to citations across video, user-generated discussions, affiliates, and blogs.
That leads to a sharper link-building question:
Is this placement only a link, or does the page provide retrievable evidence for a question our buyer actually asks?
Training data is not the practical target
A lot of AI SEO advice talks as if marketers can rewrite a model’s memory on demand. That is not the controllable part.
Smith distinguishes the core model from retrieval: an answer system can search, collect documents, and summarize what it finds. He argues that influencing the core model is extremely hard, while retrieval is the more practical surface.
That matters because retrieval works with documents that exist now:
- A useful comparison page
- A specific integration guide
- A practitioner’s video transcript
- A credible community discussion
- A third-party listicle
- A case study with clear evidence
A backlink on one of those pages is not automatically an AI citation. But the page can become part of the source environment from which an answer is assembled.
The dual-value backlink framework
Evaluate every prospective placement across five dimensions.
1. Authority value
Does the source page and domain have a legitimate reason to confer trust or discovery value?
Check:
- The site is a real business or publisher.
- The page is indexed and maintained.
- The topic matches the destination.
- The site is not built mainly to sell or exchange links.
- The page already earns traffic, links, or meaningful engagement.
Authority without relevance is a weak shortcut.
2. Retrieval value
Could the source page be retrieved for a commercially meaningful question?
Start with the question, not the channel. For example:
- Which backlink exchange platforms match businesses by niche?
- How can a SaaS company earn links without cold outreach?
- What is the difference between direct and ABC link exchanges?
- Which link-building methods avoid per-link fees?
Then inspect whether the source page answers that question clearly. A generic homepage mention offers less retrieval value than a focused comparison paragraph that explains who the product is for and how it works.
3. Evidence value
Does the passage explain why the brand belongs in the answer?
Compare these two mentions:
“Other tools include Distribb.”
versus:
“Distribb runs a managed ABC backlink exchange that matches real businesses by niche and places contextual links inside articles, avoiding immediate direct swaps.”
The second passage gives a retrieval system facts it can use. The first offers only a name.
This is not permission to stuff every feature into an anchor paragraph. The evidence must be accurate, concise, and useful to the reader.
4. Audience value
Would a real person benefit if Google and ChatGPT disappeared tomorrow?
This is the cleanest anti-spam test. Strong placements help someone make a decision, learn a process, compare options, or find a missing resource.
Smith makes a similar point about community participation. He argues for useful, authentic comments from real people and recommends disclosing who you are and where you work, then contributing useful information.
Five honest contributions can be more valuable than thousands of synthetic mentions.
5. Measurement value
Can you define what success would look like before the placement goes live?
Possible measures include:
- The source page gets indexed.
- The destination receives referral visits.
- The brand appears more often for a tracked question set.
- The source page begins appearing among cited documents.
- Assisted trial or demo conversions improve.
- Rankings move across a relevant keyword cluster.
Do not attribute every movement to one link. The goal is to create a repeatable test, not a convenient story.
Build a citation-surface map before outreach
Smith recommends mining questions from sales calls, customer support, and communities. Those are the prompts buyers are likely to bring to AI systems as well.
For each important question, record the current source surfaces:
- Owned pages: Product pages, integration pages, documentation, guides
- Editorial blogs: Tutorials, comparisons, market explainers
- Listicles: Best tools, alternatives, industry resource lists
- Video: YouTube tutorials, interviews, product demonstrations
- Communities: Reddit, Quora, specialist forums, professional groups
- Affiliates and partners: Reviews, implementation guides, case studies
Smith describes starting with target questions and identifying who currently appears as citations. That is more precise than sending a generic guest-post pitch to any domain with a high metric.
A placement rubric for AI-search-era backlinks
Score each potential page from 0 to 2 on these questions:
- Question match: Does the page answer a question your buyer asks?
- Source credibility: Is the author or publisher believable on this topic?
- Passage specificity: Can the mention include an accurate reason to consider you?
- Editorial independence: Can the publisher reject, edit, or qualify the claim?
- Human usefulness: Does the link improve the reader’s next step?
- Retrieval visibility: Does this type of source appear in current answers or search results?
- Measurement: Can you track the page and the target question after publication?
Reject low scores even when the domain metric looks attractive.
Why listicles are unusually important
A listicle often compresses a buying decision into a retrievable format:
- A clear category
- A set of named options
- Selection criteria
- Strengths and limitations
- Use-case segmentation
- Direct links to products
That structure is useful to humans, search engines, and answer systems.
But a weak listicle is just an inventory. A strong one explains why each option belongs, who should choose it, and what trade-off exists.
If you want to be added to a ranking listicle, do not send “Please add us.” Send the missing evidence:
- The exact user type you serve
- The specific job you handle
- A current feature or workflow the list omitted
- A verifiable product page or documentation source
- A concise limitation so the recommendation stays credible
Run a 30-day retrieval experiment
Smith warns that many AEO best practices are repeated without analysis. He recommends using test and control question groups.
Here is a practical version.
Week 1: Define the question set
Choose 20 to 50 commercially meaningful questions from sales, support, search queries, and community discussions.
Split them into:
- Test group: Questions where you will improve third-party evidence.
- Control group: Similar questions where you will make no off-site change.
Record current brand mentions, cited URLs, rankings, and referral traffic.
Week 2: Audit current citations
For each test question:
- Identify the pages and videos that appear repeatedly.
- Classify the source type.
- Note what facts the source uses to recommend a company.
- Find evidence gaps your brand can honestly fill.
Week 3: Create a small number of strong placements
Prioritize context-rich placements over volume:
- A comparison section with accurate selection criteria
- A case study that documents the process and result
- A listicle addition with a specific use case
- A practitioner article that cites an original resource
- A video segment that answers a narrow buyer question
Keep brand disclosure clear. Do not manufacture community endorsements.
Week 4: Measure without forcing a conclusion
Compare test and control groups:
- Did the new pages index?
- Did any tracked answers retrieve or cite them?
- Did brand inclusion change?
- Did referral traffic appear?
- Did conversions or rankings move?
- Did the control group move too?
Repeat only what produces a useful, reproducible signal.
Backlinks and AI citations are not interchangeable
A followed link can exist on a page that no answer engine retrieves. An answer engine can cite a page that uses a nofollow link or no link at all. A brand mention can influence a reader without changing either metric.
Treat these as connected but distinct outcomes:
- Backlink: A link relationship between two documents
- Search visibility: A page ranking for a query
- AI retrieval: A page selected as source material for an answer
- AI citation: A visible attribution in the generated response
- Brand inclusion: The product appears in the answer, cited or not
- Referral conversion: A person arrives and takes action
The strategy gets stronger when one high-quality page can contribute to several outcomes. It gets weaker when every task is reduced to “get a dofollow link.”
Where a backlink exchange can help
Once you know which questions matter and what a useful placement looks like, partner discovery becomes the bottleneck.
Distribb’s Backlink Exchange offers a managed way to find niche-matched, non-competing businesses and coordinate contextual links through an ABC structure. Its live page states that members can receive 3 to 5 links per month and give up to 5 links through Distribb articles.
That workflow does not guarantee an AI citation or a ranking increase. Use the framework above to evaluate each placement’s relevance, evidence, editorial quality, and measurable purpose.
Next step: Review the Distribb Backlink Exchange.
