For years, backlinks have been one of the strongest signals driving rankings in traditional search.
But have the rules changed as we enter the new era of AI search?
Together with Kevin Indig, we've run a data study to examine how backlinks influence visibility in AI-generated answers.
Let’s dive in.
Key Takeaways
TL;DR: Backlinks are important for AI visibility, but quality and authority matter far more than quantity.
- High-quality backlinks matter more than sheer volume. Domains with stronger backlink authority—measured by Semrush’s Authority Score—are mentioned more often in AI-generated answers. That said, big visibility gains only appear once a site reaches the higher authority tiers.
- Nofollow links have nearly the same impact as follow links. The study found similar correlation values for both types, with major gains appearing only in the top tiers of authority.
- Image-based backlinks can be just as powerful as text links. Once a site already has some authority, image links show even stronger correlations with AI mentions than text links—especially for higher-authority domains.
Methodology
We conducted this study in collaboration with Kevin Indig and Growth Memo.
First, we looked at 1,000 randomly selected domains inside Semrush’s AI Visibility Toolkit to understand how backlink-related signals influence visibility in AI-generated answers.
To that end, we tracked how often each domain was mentioned across five leading AI platforms:
- ChatGPT
- ChatGPT with Search
- Gemini
- Google’s AI Overviews
- Perplexity
We then compared the number of AI mentions (how many times a domain was referenced across AI-generated answers) and AI Share of Voice (a domain’s visibility in AI-generated answers relative to non-branded queries) against key backlink signals:
- Total number of backlinks
- Number of unique referring domains
- Ratio of follow vs. nofollow links
- Text links vs. image links
- Authority Score—Semrush’s composite metric that reflects backlink quality, referring domain diversity, organic traffic to linking pages, and spam signals
Finally, we ran Pearson and Spearman correlation analyses on the same dataset to examine how strongly each backlink signal was associated with AI visibility—and to identify whether those relationships were linear or showed threshold effects.
Learn more about how we measured the correlations.
Here’s what we found.
1. Authority Boosts AI Visibility, But Only Beyond a Threshold
Domains with higher Authority Scores—a metric heavily influenced by backlink quality and referring domain diversity—are more likely to appear in AI-generated answers.
However, the effect is moderate overall and only becomes significant once a domain crosses into higher authority tiers.
When we compared Authority Score to AI Share of Voice (that is, a brand’s visibility in AI-generated answers relative to non-branded queries), the results showed the following:
- Pearson: ~0.23 → a modest linear relationship
- Spearman: ~0.36 → a stronger, threshold-based signal
The gap between these two correlations reveals that backlink authority has a moderate overall influence on AI visibility, but real gains appear only after reaching certain authority levels.
In other words, incremental backlink improvements won’t move the needle much until a site reaches stronger overall authority.
“The big takeaway here is that your backlink needs to hit a minimum threshold. In simple terms: Don’t expect returns when you just get started or from linear growth. You need a minimum investment to see the expected impact.”
What This Means for You:
- Focus on breaking authority plateaus. Small link wins won’t make a visible difference. To move up, focus on earning backlinks from higher-authority, topically relevant sites, such as well-known publications, leading blogs, or industry directories.
- Expand your pool of referring domains. Getting links from a variety of different websites matters more than the sheer volume of links. Prioritize outreach to new domains in your space rather than collecting more links from the same sources.
- Balance authority with credibility. Backlinks help, but AI models also favor brands seen as trustworthy and credible. Keep content accurate, current, and well-sourced, and aim for mentions from reputable sites—even without links—to reinforce your authority.
2. Link Quality Drives AI Visibility More Than Volume
The quality of backlinks plays a major role in how often a domain appears in AI-generated answers.
When we compared Authority Score (Semrush’s proxy for link quality) with the number of AI mentions, the results showed a strong positive relationship:
- Pearson: 0.65 → strong linear correlation
- Spearman: 0.57 → strong but slightly non-linear correlation
This means domains with high-quality backlink profiles appear far more often in AI-generated answers, especially once they cross key authority thresholds.
Interestingly, the effect varies across models: ChatGPT Search and Perplexity weigh link quality the least, while regular ChatGPT (without Search on) shows the strongest correlation.
What This Means for You
- Prioritize link quality over volume. A few backlinks from authoritative, relevant websites will drive far more AI visibility than dozens of weak or unrelated ones.
- Audit your backlink profile regularly. Use tools like Semrush’s Backlink Analytics to pinpoint which domains are adding real authority, as opposed to low-value or spammy ones that could hold you back.
- Pursue meaningful authority gains. Instead of chasing small link exchanges or low-value mentions, target opportunities that genuinely boost credibility, such as digital PR, expert guest articles, or inclusion in trusted industry resources.
3. Nofollow Links Carry Almost as Much Weight as Follow Links
Nofollow links have nearly the same impact on AI visibility as standard follow links.
When we compared both (individually) to AI mentions, the correlations were almost identical.
Follow links:
- Pearson: 0.334 → moderate linear signal
- Spearman: 0.504 → strong threshold signal
Nofollow links:
- Pearson: 0.340 → moderate linear signal
- Spearman: 0.509 → strong threshold signal
This shows that AI models treat both link types similarly when evaluating authority and visibility.
The weighting also varies slightly by platform: we found that Gemini and ChatGPT tend to favor nofollow links a bit more, while Google AI Overviews and Perplexity lean more toward follow links.
What This Means for You:
- Don’t overlook nofollow links. They contribute meaningfully to AI visibility, especially when they come from high-authority sites. Remember: Getting mentioned by an authoritative and relevant source is more important than chasing quick gains.
- Watch for platform nuances and diversify your strategy. Platforms like Gemini and ChatGPT may reward strong nofollow links (e.g., from Reddit, Quora, or Wikipedia), while Google AI Overviews and Perplexity still lean slightly toward follow links.
- Keep your links—including nofollow ones—fresh and relevant. Regularly update content on pages where your brand already has nofollow mentions.
“Just like you update blog posts, update nofollow placements. Rotate in new data points, update product screenshots, or refresh stats on contributed articles your brand has earned nofollow links with, so links remain relevant and referenced.”
4. Image Links Perform as Well as (and Sometimes Better Than) Text Links
Image-based links show a stronger relationship with AI visibility than text links.
When we compared the two (individually) to AI mentions, the correlations were clear.
Image links:
- Pearson: 0.415 → stronger linear signal
- Spearman: 0.538 → strong threshold signal
Text links:
- Pearson: 0.334 → moderate linear signal
- Spearman: 0.472 → moderate threshold signal
The advantage of image links also appears once a site has built some authority.
From mid-level deciles upward, the relationship strengthens and peaks at the top, while low-authority tiers show little to no effect.
We also noticed a few differences across platforms:
- Perplexity and ChatGPT Search show the strongest link between image backlinks and AI visibility
- ChatGPT and Gemini follow with slightly weaker correlations
- Google AI Overviews shows the weakest connection, though it is still noticeable
Overall, image-based backlinks are treated as reliable reference signals across large language models—delivering the biggest visibility boosts once a domain reaches higher authority levels.
What This Means for You:
- Invest in attracting image-based links. Distribute your visuals where publishers and creators look for reference material, with easy-to-copy embed codes or source credits. Share them on platforms such as Reddit, LinkedIn, and industry newsletters where journalists and content teams source visuals.
- Focus on adding value. Create useful image assets based on original research, such as charts, infographics, product visuals, and data diagrams that others will naturally embed and link to.
“Stop treating images as accessories and start treating them as content. The best visual content explains and simplifies critical or complex topics.”
Backlinks Still Matter, But Quality Trumps Quantity in AI Search
Our study reveals a clear shift in how backlinks influence visibility: quality is king in the AI era.
Instead of trying to collect as many links as possible and to “hack” the system, you need to focus on building quality signals and breaking through authority thresholds.
Here’s what you should do:
- Prioritize meaningful authority gains over incremental link building
- Don't dismiss nofollow links from high-authority sources
- Invest in shareable visual assets that earn image-based links
- Focus on getting mentioned and cited for content that genuinely adds value
The bottom line: AI visibility comes from sustained authority, not one-off link wins. Prioritize earning high-quality backlinks that move your site into stronger authority ranges and keep your brand referenced on credible, up-to-date sources.
How we measured the correlations
To understand how backlinks and AI visibility are connected, we used two standard statistical measures: Pearson and Spearman correlations.
- Pearson checks for a linear relationship between two variables. If both values rise or fall together in a smooth pattern, Pearson picks that up clearly. But if the relationship only becomes visible after hitting a certain point (a threshold), Pearson tends to miss it.
- Spearman focuses on those threshold effects. It looks at whether the values move in the same direction, even if the curve isn’t perfectly smooth. Spearman is less sensitive to outliers and better at spotting situations where nothing happens for a while. Then the relationship suddenly strengthens after a certain level is crossed.
Both Pearson and Spearman give a score between -1 and +1:
- Closer to +1 → strong positive relationship
- Closer to -1 → strong negative relationship
- Around 0 → little to no relationship
Looking at both helped us reveal whether the relationship between various factors is linear or threshold-based.