Link building
AI Search Optimization
A commercial guide to tracking and earning visibility in AI answer engines, without invented results.
- Reading your visibility on daily trainer prompts
- Checking prompt volumes for the topic
- Reading the adidas brand kit and product lines
Across the 14 “best daily trainer” prompts, Nike is named in 88% of answers. adidas is named in 41%. The engines cite Runner’s World and Wirecutter, and you have no page that answers the question.
The topic runs at 5,400 searches a month and is climbing in the assistants. It earns a page of its own.
Your brand kit puts Supernova, not Adizero, in front of everyday runners, so the page leads with the Supernova Rise 3 and keeps Adizero for fast days. Headlines uppercase, adidas lowercase, and the one health claim is the APMA seal it holds. Staged in Documents. Nothing is published.

Run your
Fastest
Everyday.
Thanks to Dreamstrike+ cushioning the Supernova Rise 3 carries the easy miles that make up most of a training week, and keeps Adizero for the days the plan says fast. Nine weeks on our test panel sit behind every pick.
Find your daily trainer
Drafted by your AI Marketer from the Adidas brand kit and product lines. Staged in Documents, not published. Example on the sample account.
AI search optimization is the practice of tracking a business's visibility in AI answer engines like ChatGPT, Google AI Overviews, Gemini, Claude, and Perplexity, then systematically pitching for inclusion and links on the pages those engines cite. It does not guarantee rankings or traffic, but measures share of voice and identifies specific, pitchable opportunities to be cited.
What it refuses to do
It will not find you a paid placement
“This reply asks for payment in exchange for a link. I will not write that response.”
It will not say a page is unlinked when it could not read it
“Unchecked: this page returned 403 and was not read.”
It will not send from a domain of ours
“No mailbox connected. Three drafts are ready and none of them has been sent.”
On this page6 sections
What AI search optimization actually is
It is two connected practices. The first is measurement. You define the questions your buyers ask. A tool runs those questions against the AI engines on its plan, on a schedule, and records the answers. It captures the text, the brands mentioned, the websites cited, and the search queries the model used. From this, it computes visibility, share of voice, and average position for you and your competitors.
The second practice is outreach. The same tool uses its reading of the web to find specific pages that could link to you but do not yet. It finds pages that link to your competitors, resource pages, listicles, and directories in your niche. It also finds journalists asking for sources, podcasts booking guests, conferences taking speakers, and awards taking entries. For each, it drafts a personalized email for you to send from your own connected mailbox.
What it is not is a promise of placement. No ethical provider can put a link on another site. The act of sending a pitch and the act of a link appearing are separate. Optimization is the systematic work to increase the odds of the latter by improving the former.
What the market charges and how it works
Pricing follows three models. The first is software subscription. Mentionry is free for seven days on your own site, then $99 a month for its Starter plan, which tracks the answer engines and writes the pitches. The opportunity finder and drafter are part of the Starter plan; the trial writes every pitch and sends none. Enterprise pricing is quoted based on scope. These are monthly rates; an annual term is available at a discount.
The second model is per placement. Several vendors sell links on a per placement basis. They commit to delivering a specific number of links from sites in their network, often with a minimum Domain Rating. Prices range from $500 to several thousand dollars per link. The link is placed, not pitched for.
The third is retained agency services. Full service SEO and PR agencies bundle AI search visibility into monthly retainers starting around $5,000. This typically includes strategy, content creation, and manual outreach, but rarely includes the kind of automated, daily tracking of AI engines that software provides.
The risks and trade offs of each approach
Buying placements carries the highest risk. Google's link spam policy explicitly prohibits buying links that pass ranking signals. The penalty for violating this policy lands on your website, not the vendor's. A link built today can be devalued or cause a manual action tomorrow. The vendor's network is also a point of failure. If Google identifies the network, every site in it, and every site linked to from it, is at risk.
Retained services trade cost for control. You pay for human judgment and relationships. The trade off is scale and measurability. An agency cannot run your 20 key questions against five AI engines every single day and keep a permanent record of every citation. The work is often reported in summaries, not raw data. Speed is another trade off. Identifying and pitching a timely journalist query requires software that reads the web constantly.
Software subscriptions trade hands on work for precision. You operate the tool yourself. The trade off is time. You must review opportunities and send emails. The benefit is specificity. Every opportunity is judged individually by a model that can reject it if your business cannot honestly answer the query. The drafting also refuses unethical requests, like declining to pay for a link. Nothing is promised, but every pitch sent is verified later to see if a link appeared.
How to judge a provider
First, ask what is measured. If they talk about tracking ChatGPT, ask for a sample report. It should show the exact answer text, the citation URLs, and the date for a specific prompt. Avoid providers that sell estimates of prompt volume. That requires a licensed opt in panel, which almost no one has. They are measuring the world's conversations, not your visibility.
Second, ask how opportunities are found. A credible provider will name its sources. For Mentionry, those are competitor backlink analysis, direct niche searches for resource pages, and public databases like confs.tech for speaking gigs and Apple's podcast index. It does not log into HARO, Source of Sources, or Connectively, but you can paste queries from those services in. Be wary of any provider that claims a proprietary, secret database of opportunities. It likely does not exist.
Third, examine the sending mechanism. The only safe way to send outreach is from your own email address, through your own domain, using your sending reputation. This ensures replies come to you. Avoid any service that sends pitches from its own domain or a pool of addresses. It damages deliverability and you lose control of the conversation.
Finally, check the verification. After a pitch is sent, does the tool check back to see if a link was placed? Mentionry checks every pitched opening. It records the page, the anchor text, and whether the link is still there or was later removed or turned nofollow. A link that disappears is reported as a change. This is how you know what was earned, not just what was emailed.
What Mentionry does differently
It separates measurement from outreach. The Starter plan tracks ChatGPT, Gemini and Google AI Overviews every day. It runs your prompts daily, stores every answer, and calculates visibility. This is a pure measurement product. The Starter plan includes the Mine and Earn engines that find pitchable opportunities and draft the emails. This is a pure outreach product. You can buy one without the other.
It is built on a filtered feed, not a database. The system reads the web live. It found 110 candidate pages for one domain in a test. A judgment model reviewed those 110, kept 40 as plausible, and after a final check for pitchability, presented 14. The value is in the filtering, not the volume. A page that cannot be opened is reported as unchecked, never as an unlinked opportunity.
It requires no keys or connectors for discovery. You connect only your mailbox to send emails. The tool does not ask for Google Search Console access, API keys, or social logins. All discovery runs keyless. This limits some functions, like checking if a site already links to you, but it simplifies setup and reduces security risk.
It is honest about what is not built. There is no broken link builder, no relationship inventory, and no anchor text audit. It cannot see what other people type into ChatGPT. It does not backfill historical data. Recording starts the day you add a prompt. A day when every run failed shows as a gap in the chart, not as zero visibility. A brand never mentioned has no average position.
Who this is wrong for
AI search optimization is wrong for anyone who needs immediate, guaranteed links. Nothing here is promised. If you must have a link on a specific site by a specific date, buy an advertisement or a sponsored post. This is a process of consistent pitching and tracking over time.
It is wrong for businesses that cannot be a genuine source. The judgment model will reject opportunities where your business cannot honestly provide what is asked. If you sell a commodity product and a journalist asks for expert commentary, the tool will advise against pitching.
It is wrong for those unwilling to send emails. The drafts live in your browser and the emails send from your mailbox. You must review and click send. If you want a fully hands off service where an account manager does everything, a retained agency is a better fit, though it will cost significantly more.
Finally, it is wrong if you only care about traditional Google search. This channel is specifically for AI answer engines. While some opportunities may also benefit organic search, the primary goal is visibility in ChatGPT, Gemini, and Google AI Overviews. If that is not your target, this is the wrong tool.
Questions people actually ask
Mentionry tracks the engines on your plan. Available engines include ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Claude, and Perplexity. The Enterprise plan also lists Microsoft Copilot, Grok and DeepSeek, which cannot run yet. The specific engines you can measure depend on your subscription tier.